Proceedings Volume 6786

MIPPR 2007: Automatic Target Recognition and Image Analysis; and Multispectral Image Acquisition

Tianxu Zhang, Carl Anthony Nardell, Duane D. Smith, et al.
cover
Proceedings Volume 6786

MIPPR 2007: Automatic Target Recognition and Image Analysis; and Multispectral Image Acquisition

Tianxu Zhang, Carl Anthony Nardell, Duane D. Smith, et al.
View the digital version of this volume at SPIE Digital Libarary.

Volume Details

Date Published: 15 November 2007
Contents: 2 Sessions, 199 Papers, 0 Presentations
Conference: International Symposium on Multispectral Image Processing and Pattern Recognition 2007
Volume Number: 6786

Table of Contents

icon_mobile_dropdown

Table of Contents

All links to SPIE Proceedings will open in the SPIE Digital Library. external link icon
View Session icon_mobile_dropdown
  • Automatic Target Recognition and Image Analysis
  • Multispectral Image Acquisition
Automatic Target Recognition and Image Analysis
icon_mobile_dropdown
Human recognition at a distance
Recognizing people at a distance is challenging from various considerations, including sensing, robust processing algorithms, changing environmental conditions and fusing multiple modalities. This paper considers face, side face, gait and ear and their possible fusion for human recognition. It presents an overview of some of the techniques that we have developed for (a) super-resolution-based face recognition in video, (b) gait-based recognition in video, (c) fusion of super-resolved side face and gait in video, (d) ear recognition in color/range images, and (e) fusion performance prediction and validation. It presents various real-world examples to illustrate the ideas and points out the relative merits of the approaches that are discussed.
Tracking face poses toward meeting video analysis
Ligeng Dong, Linmi Tao, Guangyou Xu
We perform face tracking and pose estimation jointly within a mixed-state particle filter framework. Previous methods often used generative appearance models and naive prior state transition. We propose to use discriminating models, Adaboosted face detectors, to both measure observations and provide information for the proposal distribution which is combined with detection responses and prior transition model. Due to pose continuity, faces between discrete poses can be detected by neighboring pose-specific detectors and serve as importance samples. Thus continuous poses are obtained instead of discrete poses. Experiments show that our method is robust to large location and pose changes, partial occlusions and expressions.
Real-time video denoising arithmetic based on adaptive multi-layers background
Xiaoyun Guo, Weiming Shen, Qiong Wu, et al.
A real-time video denoising algorithm based on adaptive multi-layers background is presented in this paper. The adaptive multi-layers background, which is aimed at the demand for video denoising, not only applies to static scene, but also applies to unstable scene. The modification of multi-layers background would be accomplished by a short-term adjustment after the scene changed. Based on the multi-layers background, the video denoising algorithm promotes the distinct vision of static scene and the short-term inactive region. The model of multi-layers background adjusts step by step adaptively. So it is not in need of a long-term delay and expensive computation. Experiments demonstrate the effectiveness of this algorithm.
Extraction of residential information from high-spatial resolution image integrated with upscaling methods and object multi-features
Lixin Dong, Bingfang Wu
Monitoring residential areas at a regional scale, and even at a global scale, has become an increasingly important topic. However, extraction of residential information was still a difficulty and challenging task, such as multiple usable data selection and automatic or semi-automatic techniques. In metropolitan area, such as Beijing, urban sprawl has brought enormous pressure on rural and natural environments. Given a case study, a new strategy of extracting of residential information integrating the upscaling methods and object multi-features was introduced in high resolution SPOT fused image. Multi-resolution dataset were built using upscaling methods, and optimal resolution image was selected by semi-variance analysis approach. Relevant optimal spatial resolution images were adopted for different type of residential area (city, town and rural residence). Secondly, object multi-features, including spectral information, generic shape features, class related features, and new computed features, were introduced. An efficient decision tree and Class Semantic Representation were set up based on object multi-features. And different classes of residential area were extracted from multi-resolution image. Afterwards, further discussion and comparison about improving the efficiency and accuracy of classification with the proposed approach were presented. The results showed that the optimal resolution image selected by upscaling and semi-variance method successfully decreased the heterogeneous, smoothed the noise influence, decreased computational, storage burdens and improved classification efficiency in high spatial resolution image. The Class Semantic Representation and decision tree based on object multi-features improved the overall accuracy and diminished the 'salt and pepper effect'. The new image analysis approach offered a satisfactory solution for extracting residential information quickly and efficiently.
Framework for feature selection for cast shadow removal
Guilin Zhang, Ying Chu, Song Tian, et al.
Cast shadow cause serious problem in the extracting of moving objects because shadow pixels are liable to be misclassified as foreground. Many methods of cast shadow removal have been proposed and many features are selected in these methods. But since, moving object (MO) and cast shadow are classified by a single linear classifier. As it is known, each feature has its strength and weakness and is particularly applicable for handling a certain type of variation. In this paper, a novel framework for feature selection for cast shadow removal based on AdaBoost is proposed. Experiments are conducted on many scenes and the results prove the validation of the proposed method.
Dynamic object recognition and tracking of mobile robot by monocular vision
Lei Liu, Yongji Wang
Monocular Vision is widely used in mobile robot's motion control for its simple structure and easy using. An integrated description to distinguish and tracking the specified color targets dynamically and precisely by the Monocular Vision based on the imaging principle is the major topic of the paper. The mainline is accordance with the mechanisms of visual processing strictly, including the pretreatment and recognition processes. Specially, the color models are utilized to decrease the influence of the illumination in the paper. Some applied algorithms based on the practical application are used for image segmentation and clustering. After recognizing the target, however the monocular camera can't get depth information directly, 3D Reconstruction Principle is used to calculate the distance and direction from robot to target. To emend monocular camera reading, the laser is used after vision measuring. At last, a vision servo system is designed to realize the robot's dynamic tracking to the moving target.
Joint deconvolution of adaptive optics images from slope measurements of wavefront sensor
Bo Chen, Ze-xun Geng, Yang Yang, et al.
The atmospheric turbulence severely limits the angular resolution of ground based telescopes. When using Adaptive Optics (AO) compensation, the wavefront sensor data permit the estimation of the residual PSF. Yet, this estimation is imperfect, and a deconvolution is required for reaching the diffraction limit. It is a powerful and low-cost high-resolution imaging technique designed to compensate for the image degradation due to atmospheric turbulence. A joint deconvolution method based on slope measurements for AO image is presented. It deduces from a Bayesian framework in the context of imaging through turbulence with adaptive optics. It takes into account the noise in the images and in the Hartmann-Shack wavefront sensor measurements and the available a priori information on the object to be restored as well as on the wave fronts. Deconvolution results are presented for experimental data.
Filtering strategy on affine invariant feature detecting based on information content and distribution constraints
Liang Cheng, Jianya Gong, Peng Han, et al.
Stereo matching is one of the most important and challenging subjects in computer vision, digital photogrammetry, and image understanding. For the purpose of wide-baseline stereo matching, a novel approach on high-quality affine invariant feature extraction is proposed. The key contribution of the novel approach is a filtering strategy for affine invariant features detecting based on information content and spatial dispersion quality constraints. The essential idea is to remove the features with low information content and bad distribution, just select the high-quality features (high information content and good distribution). Based on the filtering strategy, an automatic algorithm on high-quality affine invariant feature extraction is introduced. The experiment using image sequences with different texture conditions proves that our algorithm can get much higher repeatability than the other algorithms, which is more suitable for subsequent wide baseline stereo matching.
Feature-based nonrigid image registration using multi-class Hausdorff fractions
Xiaoming Peng, Wufan Chen, Qian Ma
In a previous paper (Ref. 9) we presented a feature-based nonrigid image registration method using a Hausdorff distance based matching measure. One limitation of the method is that it is likely to fail in "ambiguous" cases where a part of the features in the source image are nearer to a prominent number of non-corresponding features in the target image than to their corresponding ones. To partly alleviate this limitation, in this paper we propose a new feature-based nonrigid image method that uses multi-class-Hausdorff-fractions-based similarity matching measure. We first divide features into a finite number of classes, then we calculate a similarity matching measure by adding up the forward and backward multi-class Hausdorff fractions of the classes. The new similarity matching measure outperforms that used in our previous work, given that the features in the images to be registered can be correctly classified. We also adapted the optimization procedure of our previous method so that it can work appropriately with the new similarity matching measure. The new method, introducing only a small computational load, is capable of reducing undesired matching of features that are adjacent to each other but belong to different classes.
Object region extraction based on graph cut and application in image retrieval
Li Guo, Lingjun Wang, Xinghua Sun, et al.
This paper introduces the technique of graph cut into the extraction of object region and applies the corresponding result of object region extraction into the image retrieval based on object region. The main idea of image retrieval based on object region is to use the feature of object region instead of the feature of global image to participate in the image retrieval. In the field of graphics there is a technique called graph cut, which can be used to figure out the contour of object under the interaction of users. The graph cut algorithm can be used to verify the correctness of object region extraction, and the users' input about seeds can be simulated according to the initial object region extracted. The usage of graph cut can make the object region extracted more precisely and thus the performance of image retrieval based on object region can be improved. Experiments show that the object region extraction algorithm based on graph cut is valid and the subsequent image retrieval results accord with the human visual perception much more than the ones without the usage of graph cut.
Improved algorithm for image decomposition based on bidimensional EMD
Jingbo He, Fuyuan Peng
According to the system intrinsic quality of self-comparability and the empirical mode decomposition algorithm of completeness and stability, an improvement algorithm for EMD image decomposition is presented. It is integrity, fast and effective. Some aspects were improved that bidimensional interpolation methods and end conditions of getting intrinsic mode image. Three questions of general algorithm for EMD image decomposition were solved. First, the algorithm for image decomposition was slow; Second, some points were not contained because of delaunay triangulation; Third, the end condition of algorithm was not accuracy. Experiments were made by Matlab and the validity of the improvement algorithm was validated.
Algorithm of image fusion based on finite ridgelet transform
Kun Liu, Lei Guo, Weiwei Chang, et al.
Finite ridgelet transform (FRIT) overcomes the weakness of wavelet transform representing in two or higher dimensions and FRIT can efficiently represent the singularity of linear in image. When FRIT is applied to image fusion, the characters of original images can be effectively extracted and more important information is preserved. In this paper, we discussed the advantage of FRIT applying to the image fusion and the process of the fusion algorithm based on FRIT in details. Two sets of images are taken as experimental data, subjective and objective standard are used to evaluate the results. The experiment results show that the FRIT algorithm gets much better fusion results than wavelet, and image fusion algorithm based on FRIT is an effective and feasible algorithm.
Extraction of non-structured object contours based on multi-scale entropy difference operator
Li Liu, Fuyuan Peng, Kun Zhao, et al.
Contour information is regarded as important characteristic in computer vision. It is difficult to extract Contour information from the non-structural object due to its complicated structure. This paper present a novel concept of Multi- Scale Entropy (MSE) based on traditional Entropy that can be used to perform reliable extracting contours from non-structured objects such as smoking and rocks. The variety of image information amount was presented dynamically by this means. The Multi-Scale Entropy Difference (MSED) can present the break part of the image gray information and recognize the boundary of object and background effectively. Finally the non-structural object contours was extracted by Maximal Multi-Scale Entropy Difference (MMSED). Experiments have shown that the operator can extract stable contours from non-structural objects and eliminate the interior complex texture structure effectively.
Point matching of two images under large different resolution
Image matching is the first step in almost any 3D computer vision task, and hence has received extensive attention. In this paper, the problem is addressed from a novel perspective, which is different from the classic stereo matching paradigm. Two images with different resolutions, that is high resolution versus low resolution are matched. Since the high resolution image only corresponds to a small region of the low resolution one, the matching task therefore consists in finding a small region in the low resolution image that can be assigned to the whole high resolution image under the plane similarity transformation, which can be defined by the local area correlation coefficient to match the interest points and rectified by similarity transform. Experiment shows that our matching algorithm can be used for scale changing up to a factor of 6. And it is successful to deal with the point matching between two images under large scale.
Image inpainting under different displacement view images
In this paper, Image inpainting based on different displacement view images is proposed, which is the problem of filling in the occluded or damaged regions of an image by the visible information from other different displacement view images. The key problems are how to convert the visible information in different displacement view images to consistence and to use available information to repair the target image. The first step of our method is to divide all images into different scene planar regions and to transform all image regions into current view by projective transformation computed from the matched points; thus the visible information can be directly used, then a new inpainting algorithm based on image fusion with spatial frequency is applied. The experiment shows good and harmonious results for the repaired image.
Method of airport detection based on multi-resolution and multi-spectrum remote sensing images
Cheng Wang, Hai Sun, Bo Ren
Airport detection is one of the most important research topics in the remote sensing application field. In the field, majority of airport detection methods are designed based on different single resolution optical remote sensing images. However, because of the RS data's limitation and algorithms' shortages, these methods are not likely to have a perfect performance; inaccuracy and inefficiency are the two main drawbacks. In recent years, while much research of airport recognition has been done on single resolution images, less work has been tried to study on multi-resolution and multi-spectrum images together. Nevertheless, after many experiences, the latter method is proved available and attemptable in our research. Based on the different characteristics of multi-resolution and multi-spectrum images, we try to design a synthetical model which includes algorithms of structure feature analysis and spectral analysis. In the model, different advantages of images and methods are combined to reduce influence of each other's limitations. Finally, an airport-detection system is implemented based on the synthesis method. This paper focuses on the designment and implementation of the synthesis model, enrich the research of airport recognition and solve many problems. Section 1 provides an overview of the synthesis model. Section 2 discusses the core algorithms of the method in detail. Section 3 demonstrates airport detention system and Section 4 concludes the paper.
Quadratic programming based multi-target tracking in infrared image sequence
Jiuqing Wan, Jinsong Yu, Xiaoqing Zhang
In this paper a novel approach for multiple maneuvering targets tracking in infrared image sequence is proposed. For each target, the measurement weights and model weights, which are essential for data association and multi-model based filtering, are determined jointly by quadratic programming. The proposed method avoids events enumeration process in traditional tracking algorithm such as JPDA and hill-climbing optimization process in classical EM-based algorithm. Simulation results show that compared with IMM+JPDA and recursive EM based algorithm, the proposed quadratic programming based algorithm can achieve good performance in both tracking accuracy and computational complexity.
Adaptive watermarking algorithm based on wavelet transform
An adaptive watermarking method based on discrete wavelet transform is proposed in this paper. To get good imperceptivity and robustness, the energy of sub-blocks and the texture of the carrier image have been used to determine where the watermark will be embedded. And the watermark embedding intensity will be determined by the noise visibility function. Experimental results show that this method is effective.
SAR image edge detection combining radon transform with RDWT
Xinyan Yang, Licheng Jiao, Jiajun Wang
The Synthetic Aperture Radar (SAR) is a promising sensor to obtain high-resolution images. But the presence of speckled noise brings many difficulties to SAR image processing, especially in edge detection. The edge detection methods available cannot solve this problem perfectly. An effective edge detection algorithm based on the directional information is presented, in which, two transforms are introduced, one is Fast Slant Stack transform, a new radon transform with the advantage of speed and invertibility, the other is Redundant Discrete Wavelet Transform(RDWT) whose best performance in edge detection is shift-variance characteristic. Besides, overlapped windows and soft threshold utilized to reduce the wrong detection probability and improve the location precision. Finally, comparisons with related methods are given in this paper. Experimental results prove the proposed algorithm can obtain valid edge information in SAR images.
Underwater image modeling using multi-scale fractional lévy stable motion
Xutao Li, Lianwen Jin, Fuyuan Peng, et al.
Fractal-based analysis provides an excellent explanation of the ruggedness of natural surface. Fractal-based description of image texture has been used effectively in characterization of natural scene. However, a real surface is not always so perfect that keep invariable self-similarity in whole scale space and it seems to be multi-scale. The present work focused on the modeling of underwater image. We analyzed the increment distribution and the self-similarity of several typical objects, and the results show that the traditional FBM model is not suitable for modeling such objects. And then one kind of self-similar stable process, Fractional Lévy Stable Motion (FLSM) is discussed. Based on FLSM, we proposed a new model, Multi-scale Fractional Lévy Stable Motion (MFLSM) in which the self-similarity parameter H (Δs) is a variable with respect to measure scale s. Furthermore such 2-D model is applied to model the scattering image. The simulation result shows that MFLSM represents the multi-scale self-similarity and the speckle of underwater optical imaging.
Color image filtering based on impulsive noise detection
Dehua Li, Lianghai Jin, Xingzhong Yao, et al.
This paper introduces a new class of switching vector median filter. The proposed algorithm first uses four directional masks to analyze the color difference between the central pixel and its neighborhood pixels in the RGB color space and classify each color pixel into noisy pixel or noise-free one, and then employs the standard vector median filtering operations in the detected noisy locations to restore the corrupted pixels and leave the noise-free ones unchanged. The simulation results show that the proposed method excellently suppresses impulsive noise as well as preserving the image details well, and significantly outperforms the existing vector filtering solutions in terms of both the objective measures and the perceptual visual quality.
Automated airport targets detection and location based on radar imagery
Yongshe Shao, Ying Chen, Jing Li
Because of the influence of speckle noise, it is difficult to extract the edge of the targets and locate the airport automatically in the low SNR real aperture radar imagery by the traditional ways. The paper presents a kind of algorithms that could improve the dependability of the image matching and locate the airport target in the real aperture radar images. At first, the original real aperture radar imagery is enhanced on fuzzy property domain, which makes the airport target more prominent. Then, the airport runway is detected using the Radon transform, and the end points and width of the airport runway are identified using the correlative knowledge of the airport. At last, the airport runway extracted is processed and the airport target is located with the wavelet transform and the least square image matching. The experiment shows that the method could detect and locate the airport target well.
New method of cloud synthesis and application in image segmentation
Kai Xu, Kun Qin, Deren Li
Currently, cloud model has been successfully used in data mining, digital watermark and other fields. This paper proposes a new method of cloud synthesis which is the key technology of cloud model. Based on the new method, a novel image segmentation algorithm is proposed. Experiments show that the new method of cloud synthesis is more effective and appropriate for image segmentation, and the novel image segmentation algorithm is more effective and robust than the traditional image segmentation algorithms.
New approach for all-star pattern recognition
Guangjun Wang, Wei Wang, Yongtao Wang, et al.
In this paper, a new star pattern recognition approach has been developed, the basic idea of which is to extract the star pattern information, to fit together the information as a pending data set, putting all reference star information into a criterion data set, and then to compute the relative Hausdorff distance between two sets. The pending star is recognized according to the minimum Hausdorff distance. The approach involves most of the information of pending star dimensional configuration. Therefore, it is strong robust for the disturbance of noise, distortion, and few meteors. In the case of different disturbances, for instance, random noise, imaging distortion, and few meteors etc, semi-physical simulation experimental result indicates that the approach is of good recognition effect.
Road information extraction from IKONOS imagery based on clustering analysis and mathematical morphology
Yang Song, Youchuan Wan, Shaohong Shen, et al.
In this paper, we present an approach based on clustering analysis and mathematical morphology to extract road information from IKONOS imagery. This road information extraction approach includes several key modules: Texture analysis based on the multi-band image to obtain two new features of "MLen/MWid" to improve the road clustering analysis; In order to optimize the primal binary imagery of road object area resulting from clustering process, a texture analysis defined on binary imagery--"BATS" is presented, which ulteriorly expel the non-road pixels from the road area binary imagery; Furthermore, we carry out the process to extract road centerline network from the binary imagery of road object area based on mathematical morphology, through the process, several other methods, such as connectivity analysis, raster to vector transform, etc., are integrated.
Detecting moving objects by background difference and frame-difference
Yihuan Zhao, Zulin Wang
The performances of Background Difference Method and Frame-difference Method for detecting moving objects are analyzed. It is found that as the second method is characterized by being sensitive only to moving objects, it can be used to update the background image, an important step for the first method. A new technique integrating both methods is then proposed. Using the consecutive frames differencing images and the detection result of the previous frame, we restore background model, which can be utilized later to eliminate the detection error caused by the Frame-difference Method. Based on this truth, we design a detecting system to realize the new algorithm. Experiments on indoor and outdoor video streams show that the new technique has strong adaptability and veracity.
Recognition method of harbor target in remote sensing image
Yongjie Huang, Shuguo Wang, Kun Xing
The research on harbor recognition from remote sensing image is a very complicated problem due to complex environment and features. On this mentioned above, a method is presented to recognize the targets with various transformations on the basis of harbor targets extraction. This paper investigates the environmental characteristics of the harbor target. After preprocessing and segmenting real-time image, the inside region which reflects the nature of the harbor is extracted. From inside region the moment invariants can be calculated that it has the invariability with displacement, rotation and scale. According to the practical application, an experimental system based on harbor targets recognition is established. The result indicates that the harbor targets in remote sensing images can be recognized accurately using the method presented.
Robust L1 PCA and application in image denoising
The so-called robust L1 PCA was introduced in our recent work [1] based on the L1 noise assumption. Due to the heavy tail characteristics of the L1 distribution, the proposed model has been proved much more robust against data outliers. In this paper, we further demonstrate how the learned robust L1 PCA model can be used to denoise image data.
Moving targets detection based on the 1-norm of the optical flow difference vector
Tongsheng Shen, Chunxin Wang, Yuye Zhang
With the development of space field, knowing and mastering the status of all kinds of space targets is necessary means to use the space resource efficiently. Moving targets detection is very important to mastering the status of the space targets by using the sequence space images since it is the precondition and basic of targets identify. Traditional method of moving targets detection which through the movement compensation is mainly faced to the linear moving of the background and seldom focus the research on the rolling moving around the center of the view field. This paper analysis the movement rule of targets and background stars in the sequence images and propose a moving targets detection method with the circumrotate moving background around the center of the view field based on the cumulate 1-norm of the optical flow difference vector. The first step is to get moving streak of targets and stars by the cumulate 1-norm of the optical flow difference vector. The second step is to find the entire moving point region by analysis moving streak and then detect targets. At the end, we examine the method by processing a sequence images, the result shows that the method is useful to the space targets detection.
Light-microscopic image restoration via edge detection
Mingzhu Sun, Xin Zhao, Guizhang Lu
A semi-blind image restoration approach is put forward in this paper for light-microscopy system. Microscope creates unavoidable light artefacts because of the Point Spread Function (PSF) of the optical system, and our early research shows that PSF of the microscopy system can be modeled as isotropic Gaussian blur. That is the motivation of this paper. We present an algorithm, based on functional minimization, which integrates Canny edge detection with semi-blind deconvolution. An alternating minimization (AM) implicit iterative scheme is devised to recover the image and simultaneously identify PSF. Good performance is observed with numerically blurred images and really microscopic images, even under the presence of high noise level.
Difference-templates based target tracking method
Haibo Luo, Zelin Shi, Deqiang Li, et al.
This paper proposes a difference-templates based target tracking method (DTBTTM) with the originality of constructing a collection of difference templates that represent the varying characteristics of target region, such as translation, scale, and illumination. DTBTTM method uses the linear combination of such difference templates to represent the variation of target region, and computes coefficients with respect to the corresponding templates. The final target position and window size can be determined with these coefficients. DTBTTM method simply solves linear equations, and is quite different from correlation method in which 2-dimensional search is required to calculate similarity between pre-defined template and the region of interest. Experimental results show that the DTBTTM is highly adaptable to the variation of target region, and is robust to the variation of translation, scale, illumination, and even occlusion.
Model of motion perception based on biological vision principle
Hao Zhao, Zhengzhi Wang, Jiaomin Huang
According to the principles of biological vision, the paper proposed a model of motion perception based on Grossberg's Formotion (form-motion) BCS model for detecting moving targets and their directions in a set of image sequence. It is a parallel processing system including static flow and motion flow while the Formotion model is a serial processing from static flow to motion flow. Additionally, the model here carried out developments at several stages to make it available in complex real scenes. It imported multipoint inhibition in transient cell network, made use of Gaussian kernel inhibition of the opposite direction and designed a new cell membrane equation to obtain clear motion boundaries along the motion directions. The simulations indicate that it can detect moving objects in real scenes fast and distribute their boundaries to the corresponding motion directions exactly. It is successful to make analysis of motion image sequence by this model.
Study on quality check methods of digital orthoimages
Yawen Liu, Zhijiang Li, Zheng Ji, et al.
In this paper, we describe an effective method for fast checking the quality of digital orthoimages. In order to ensure the accuracy and precision of digital orthoimages (DOM) before putting into database, we consider metadata, coordinate data, geometry accuracy, and histogram quality of images as evaluating rules. Template matching is used in Metadata inspection. Coordinate data inspection is based on the relation between the standard file name of maps and coordinates of maps. Image matching technique is performed in geometry accuracy inspection. Image histogram quality is evaluated by the normal distributing curve characteristic. The method is tested feasible in applications for checking the quality of digital orthoimages.
Novel corner detector based on lifting wavelet transform and SUSAN algorithm
Jianhong Zhang, Ping Zhang
This paper analyzes the SUSAN algorithm and points out its three shortcomings: fixed brightness difference threshold, examining every pixel step-by-step without selection and coarse USAN area calculating method. To overcome these shortcomings, a novel corner detector is proposed. Lifting wavelet transform is used to obtain the high frequency component of the input image. Corner candidates and the adaptive brightness difference threshold are obtained from the high frequency information. Then the SUSAN algorithm is improved to select the real corners from the candidates. In the improved SUSAN algorithm, USAN area is calculated according to both the similarity of pixels' brightness and the connectivity of the pixels in the circle mask. Experiment results show that the proposed corner detector is faster and more effective than both the traditional SUSAN algorithm and the adaptive algorithm proposed in references.
FLIR image segmentation based on DA-GMRF model
In order to conquer the drawback of over-smoothness in the MRF model, a kind of discontinuity-adaptive Gaussian Markov random field (DA-GMRF) model is defined, in which the edge information of image is used to construct corresponding energy functions. After this a FLIR image segmentation method based on DA-GMRF model is proposed, which includes initialization and optimization of the label field. A multi-threshold image segmentation algorithm based on potential function of gray histogram is presented to initialize the label field. This algorithm can determine region number and multi-thresholds automatically. Metroplis Sampler algorithm is adopted to optimize the label field. Segmentation experiments on several images show that the algorithm proposed is effective.
Ship detection algorithm in SAR images based on Alpha-stable model
Changcheng Wang, Mingsheng Liao, Xiaofeng Li, et al.
This paper proposes a new ship detection algorithm based on Alpha-stable model for detection ships in the spaceborne synthetic aperture radar (SAR) images. The current operational ship detection algorithm is based on Constant False Alarm Rate (CFAR) method. The major shortcoming of this method is that it requires an appropriate model to describe statistical characteristic of background clutter. For multilook SAR images, the Gaussian model can be used. However, the Gaussian model is only valid when several radar looks are averaged. As sea clutter in SAR images shows spiky or heavy-tailed characteristics, the Gaussian model often fails to describe background sea clutter. In this study, we replace Gaussian model with Alpha-stable model, which is widely used in the application of impulsive or spiky signal processing, to describe the background sea clutter in SAR images. Similar to the typical Two-parameter CFAR algorithm based on Gaussian distribution, we move a set of local windows through the image and finds bright pixels that are statistically different than the surrounding sea clutter. Several RADARSAT-1 images are used to validate this Alpha-stable model based algorithm. The experimental results show improvements of using Alpha-stable model over the Gaussian model.
Driver fatigue alarm based on eye detection and gaze estimation
Xinghua Sun, Lu Xu, Jingyu Yang
The driver assistant system has attracted much attention as an essential component of intelligent transportation systems. One task of driver assistant system is to prevent the drivers from fatigue. For the fatigue detection it is natural that the information about eyes should be utilized. The driver fatigue can be divided into two types, one is the sleep with eyes close and another is the sleep with eyes open. Considering that the fatigue detection is related with the prior knowledge and probabilistic statistics, the dynamic Bayesian network is used as the analysis tool to perform the reasoning of fatigue. Two kinds of experiments are performed to verify the system effectiveness, one is based on the video got from the laboratory and another is based on the video got from the real driving situation. Ten persons participate in the test and the experimental result is that, in the laboratory all the fatigue events can be detected, and in the practical vehicle the detection ratio is about 85%. Experiments show that in most of situations the proposed system works and the corresponding performance is satisfying.
Image denoising with window shrink wavelet coefficients by adaptive threshold
Yifan Zhao, Jiuxian Li, Liangzheng Xia
We introduce an adaptive wavelet coefficients shrinkage method and apply it to image denoising. Donoho's denoising scheme which is based on thresholding the wavelet coefficients, eliminates too many wavelet coefficient without considering the image's local characteristics. In this paper we propose a new shrinkage method which can modify the magnitude of shrinkage by considering neighboring wavelet coefficients and variance of noise. So we can use more wavelet decomposition levels than other wavelet shrinkage methods to recover the noisy images. The proposed method outperforms the other methods given in literature, while its implementation and concept are both simple.
Fast approach to super-resolution image reconstruction
Di Zhang, Weiping Hu, Jiazhong He, et al.
Super-resolution image reconstruction produces a high-resolution image or high- resolution image sequences from a set of shifted, blurred, and decimated versions thereof, and has been proven to be extremely useful in early vision, video surveillance, and other applications. However, as magnification increases, previously published techniques get worse either in computational complexity or ringing artifacts. In this paper, a fast approach is proposed to reduce both the ringing artifacts and the computational complexity. Experiment results demonstrate that the new approach is more efficient and can provide much better reconstruction quality in comparison with normal super-resolution algorithms.
New edge detection method of SAR images
Jing Zhang, Guo-hong Wang, Zhi-yong Yang
Clear and continuous edge feature is important to analyze and interprete SAR images. In this paper, a new edge detection of SAR images is presented by analyzing regularization method. There are three steps. Firstly, using a modified regularization method reduces the speckle noise. And then, a statistical method is proposed to segment the target of interest and the shadow. At last, edge detection is realized by a window method. Comparing with traditional methods, experimental results with MSTAR dataset show that this method can maintain detail feature, resolve the broken edge and decrease noise efficiently. This algorithm has wonderful edge-detection performance.
Research on feature recognition algorithm for space target
Jian Zhang, Xiaodong Zhou
In this paper, a robust methodology on space target feature recognition is introduced. Aiming at area space target, its invariant features about geometry, affine transform and gray-level changing are extracted. Using the Backpropagation Fuzzy Neural Network (BPFNN) classifier, different models of target are classified and recognized. Aiming at point space target, firstly, local gray-level probability is computed and used to separate target and stars from background by setting threshold. Then by using multi-frame image accumulation, the contrast between target and stars is enhanced. Finally, target's accurate coordination has been achieved through centroid method with gray-level weighted. It has been improved that algorithm adopted in this study can reach approximately 93% accuracy of recognition for area target and 0.1 pixel of positioning accuracy for point target.
Deformable target tracking method based on Lie algebra
Yunpeng Liu, Zelin Shi, Guangwei Li
Conventional approaches to object tracking use area correlation, but they are difficult to solve the problem of deformation of object region during tracking. A novel target tracking method based on Lie algebra is presented. We use Gabor feature as target token, model deformation using affine Lie group, and optimize parameters directly on manifold, which can be solved by exponential mapping between Lie Group and its Lie algebra. We analyze the essence of our method and test the algorithm using real image sequences. The experimental results demonstrate that Lie algebra method outperforms other traditional algorithms in efficiency, stabilization and accuracy.
Image denoising based on local adaptive multi-scale wavelet least squares support vector regression (MWLS_SVR)
Rather than attempting to separate signal from noise in the spatial domain, it is often advantageous to work in a transform domain. Building on previous work, a novel denoising method based on local adaptive multi-scale wavelet least squares support vector regression is proposed. Investigation on real images contaminated by Gaussian noise has demonstrated that the proposed method can achieve an acceptable trade off between the noise removal and smoothing of the edges and details.
Method of passive ranging from infrared image sequence based on equivalent area
Weiping Yang, Zhenkang Shen
The information of range between missile and targets is important not only to missile controlling component, but also to automatic target recognition, so studying the technique of passive ranging from infrared images has important theoretic and practical meanings. Here we tried to get the range between guided missile and target and help to identify targets or dodge a hit. The issue of distance between missile and target is currently a hot and difficult research content. As all know, infrared imaging detector can not range so that it restricts the functions of the guided information processing system based on infrared images. In order to break through the technical puzzle, we investigated the principle of the infrared imaging, after analysing the imaging geometric relationship between the guided missile and the target, we brought forward the method of passive ranging based on equivalent area and provided mathematical analytic formulas. Validating Experiments demonstrate that the presented method has good effect, the lowest relative error can reach 10% in some circumstances.
Adaptive template-updating strategy based on singular value decomposition
The problem of image matching and target tracking based on singular value decomposition (SVD) is discussed. The SVD has robust performance that is invariant to image disturbance and it makes the singular value credible to represent the image as an algebraic feature. A template-updating strategy is proposed to update the current template based on the scale invariant character of the singular value vector. The updated template that contains the accurate target is adaptively acquired according to the singular value's scale invariance. Experiments are performed on a large test set and the results show that the proposed strategy is practical and efficient in target tracking.
Relaxation-based approach for object recognition
Tongwei Lu, Xiaoying Gao, Nong Sang
Object recognition can be formulated as matching image features to model features. When recognition is based on point feature, feature correspondence should be one-to-one. However, due to noises, repetitive structures and background clutters, features don't match one-to-one but one-to-many. By using the multi-scale feature point technique, we present an object recognition algorithm that makes features match one-to-one. First, it determines the correspondence by using the location, scale factor, orientation and local invariant descriptor of each feature point. Then a vote is recorded for the center, scale factor and rotation angle of object for each correspondence point. This approach can recognize the objects in the case of scale change, rotation angle changes and partial occlusion. Experimental results demonstrate the robustness of the overall approach on various image pairs.
New color-based tracking algorithm for joints of the upper extremities
Xiangping Wu, Daniel H. K. Chow, Xiaoxiang Zheng
To track the joints of the upper limb of stroke sufferers for rehabilitation assessment, a new tracking algorithm which utilizes a developed color-based particle filter and a novel strategy for handling occlusions is proposed in this paper. Objects are represented by their color histogram models and particle filter is introduced to track the objects within a probability framework. Kalman filter, as a local optimizer, is integrated into the sampling stage of the particle filter that steers samples to a region with high likelihood and therefore fewer samples is required. A color clustering method and anatomic constraints are used in dealing with occlusion problem. Compared with the general basic particle filtering method, the experimental results show that the new algorithm has reduced the number of samples and hence the computational consumption, and has achieved better abilities of handling complete occlusion over a few frames.
Effective Gaussian mixture learning and shadow suppression for video foreground segmentation
Robust and efficient foreground segmentation is a crucial topic in many computer vision applications. In this paper, we propose an improved method of foreground segmentation with the Gaussian mixture model (GMM) for video surveillance. The number of mixture components of GMM is estimated according to the frequency of pixel value changes, the performance of GMM can be effectively enhanced with the modified background learning and update, new Gaussian distribution generation rule and shadow detection. In order to improve the efficiency, illumination assessment is used to decide whether there are shadows in the given image. Shadow suppression will be adopted based on morphological reconstruction. Besides, the detection of sudden illumination change and background updating are also presented. Results obtained with different real-world scenarios show the robustness and efficiency of the approach.
Facet-based adaptive anisotropic diffusion for image selective smoothing
In each step of anisotropic diffusion smoothing, noises must be managed to get better results. The mostly used method is Gaussian filtering. However, the standard deviation of the Gaussian filter can't be accurately obtained and it should change during the iterative process. Another problem is how to select a proper standard deviation to reducing noises while preserving edges. Actually, facet model fitting can be taken as a natural way to overcome the drawbacks mentioned above. Facet model fitting has the low-pass filtering performance adaptive to the image during evolution of diffusion; it can also achieve balanced results for noise reduction and edge preserving. Experiments show the method can preserve more edges as well as obtain higher peak signal-to-noise ratio as compared to other anisotropic diffusion based selective smoothing approaches.
Maximal probability method of boundary extraction based on particle motion
Liantang Lou, Zhongliang Fu, Si Jiang
To overcome the main drawbacks of global minimal for active contour models (L. D. Cohen and Ron Kimmel) that the contour is only extracted partially for low SNR images, we present a new boundary extraction method, called maximal probability method of boundary extraction. We extend the description of boundary extraction from the point of view of classic mechanics to quantum mechanics, and propose a new boundary extraction approach based on maximal probability of a moving particle from one point to another. Our method is based on finding a path of maximal probability. The method includes four sequential parts: Explain boundary extraction from quantum mechanics; Estimate the probability that a particle moves from a point to another; Find a path of maximal probability between two points; Extract closed boundary from a single point by dividing the image into two small images. We show examples of our method applied to real images to compare our method with global minimal for active contour models. The experiments demonstrate that our method can overcome the main drawbacks of global minimal for active contour models.
Real-time target tracking with particle filter in moving monocular camera
Guocheng Liu, Yongji Wang
In this paper, we propose an improved particle filter algorithm for real-time tracking a randomly moving target in dynamic environment with a moving monocular camera. For making the tracking task robustly and effectively, color histogram based target model is integrated into particle filter algorithm. Bhattacharyya distance is used to weight samples by calculating each sample's histogram with a specified target model and it makes the measurement matching and samples' weight updating more reasonable. In order to reduce sample depletion, the improved algorithm will be able to take the latest observation into account. The experimental results confirm that the method is effective even when the monocular camera is moving and the target object is partially occluded in a clutter background.
Target trace acquisition method in serial star images of moving background
Biao Chen, Chunhua Zhang, Xiaodong Zhou
The paper put forward a small target trace acquisition algorithm in serial star images of moving background. For the CCD serial star images, while the space target is moving, the background is still moving because of the CCD flat is moving. At the same time, because of dithering of CCD flat and some other reasons, the gray value of target is not invariable, so much as to disappear in the background noise. The algorithm can detect the discontinuous trace of small space target with high accuracy. Originally, the K-sigma cross projection method is used in background star and target area acquisition. Then, a group of N brightest stars in every image is used to estimate the moving vector of the whole background. Finally, a target trace acquisition method is put forward. The simulation experiment is done with two small targets in 16 frame serial star images, and results of trace acquisition show that the algorithm can detect traces of the two targets with high accuracy.
PCA based forward-looking infrared airport recognition combining intensity and shape feature
Wei Liu, Jinwen Tian, Xinwu Chen
In this paper, a novel method based on PCA with shape and intensity information is proposed for infrared forward-looking airport recognition. Here, PCA is used to perform feature transformation and airport recognition. It maps an input image into a low-dimensional feature space in order to make the mapped features linearly separable. And the input image of conventional method only uses intensity information. The proposed method not only considers the intensity but also adopts shape-mask to emphasize the important object area information. The novel method is evaluated based on the sequence of infrared forward-looking airport images by using different airport recognition methods such as BP networking and SVM. The experiment's results have been compared based on percentage of correct classification, computation complexity and amount of training data, which show that this new method is superior to other recognition approach on computation complexity under almost the same recognition accuracy.
Linear object extraction from high-resolution image based on MRF and Bayesian
Changhui Yu, Lingkui Meng, Yaohua Yi, et al.
The commercialization of high resolution remote sensing image provides the image application more widely space. How to extract the interested important objects quickly and exactly from remote sensing image is always the research focus. After analyzed the characteristics of road in high resolution image, the paper constructed the road extraction model based on MRF and Bayesian. And finally the validity of the method was confirmed by an example.
Multiscale image segmentation based on one class SVM and wavelet
HaiLin Xiang, Qiang Sun, LiCheng Jiao
A supervised multiscale image segmentation method is presented based on one class support vector machine (OCSVM) and wavelet transformation. Wavelet coefficients of training images in the same directions at different scale are organized into tree-type data as training samples for OCSVMs. Likelihood probabilities for observations of segmentation image can be obtained from trained OCSVMs. Maximum likelihood classification is used for image raw segmentation. Bayesian rule is then used for pixel level segmentation by fusing raw segmentation result. In experiments, synthetic mosaic image, aerial image and SAR image were selected to evaluate the performance of the method, and the segmentation results were compared with presented hidden Markov tree segmentation method based on EM algorithm. For synthetic mosaic texture images, miss-classed probability was given as the evaluation to segmentation result. The experiment showed the method has better segmentation performance and more flexibility in real application compared with wavelet hidden Markov tree segmentation.
New change detection tracking method based on background model
Ruolan Hu, Xiao Zhou, Tao Zhang, et al.
Background constructing and updating are the two key problems for the background subtraction method used to the detection of the moving target. In the paper, confident exponent is proposed to construct the original background, and an adaptive strategy is applied to update the background in time. This method was applied on tracking the car in the exam of driver's license. The experiments showed that this method was very promising and could overcome the background change and target change.
Realization of shot boundary detection based on a unique classifying method
Meng Cai, Kefeng Zhang, Fan Xu, et al.
A new shot boundary detection system based on a unique classifying method was proposed. The pixel matched difference algorithm was improved and two new features: pixel similarity degree and pixel mean difference were proposed. Also a new classifying method for shot boundary transition, including cut, fade out or in, short gradual transition, long gradual transition and other gradual transition was proposed. These improved features showed good performance in the detection; the algorithm based on classifying method traded off well between performance and cost, both of which were proved by the experiment.
Infrared image object recognition based on invariant contourlet sub-band features
Xue Mei, Liangzheng Xia, Jinguo Lin
A novel feature descriptor-contourlet Fourier invariant feature, which combine contourlet decomposition and Fourier transforms and is translation-, rotation-, and scale-invariant, is put forward in this paper. Firstly, the translation and rotation invariant are achieved by Fourier transform along the circles that around the mass center of the scale-normalized target. Then statistic parameters of General Gaussian density (GGD) model of each contourlet sub-bands are evaluated. GGD parameters and contourlet decomposition coefficients are both as the features, which not only with rotation, shift and scaling invariant, but also with the contourlet inherent property of multi-resolution, local and multi-direction. We present experimental results using this descriptor in infrared image recognition, and it shows this descriptor is a good choice for object recognition.
New texture segmentation approach based on multiresoluton MRFs with variable weighting parameters in wavelet domain
In this paper, a new texture segmentation method based on MRMRF in wavelet domain is proposed. Unlike the common used MRMRF, a variable weighting parameter is employed to combine the feature field and the labeling field, which makes the two random fields dominant successively in the procedure of image segmentation and makes it possible to get a more accurate segmentation result. Experimental results demonstrate that the proposed method is effective for texture image segmentation.
Image segmentation using improved watershed transform
Geng Li, Xiaoqian Chen
Watershed transform is a morphological image segmentation method which is widely used in medical image processing, video processing and many other fields. The ultimate goal of the transform is to identify the objects of interest in the input image. However, the main problem of watershed transform is its sensitivity to intensity variations, resulting in over segmentation problem. The main goal of this work is to overcome the drawback of the traditional watershed algorithm. The proposed algorithm in this paper is to retain the most significant regions of the image when the image is under noisy conditions and the objects in the image have detailed textures.
Application of SGNN-based method in image segmentation
Lu Li, Hong Jiang, Zhang Ren, et al.
In this paper, a SGNN (Self-Generating Neural Network)-based method is applied to image segmentation, which is implemented automatically by autonomously clustering the pixels according to their gray values. The optimization of SGNN is studied to further improve the accuracy and robustness, as well as to reduce the computational complexity of the segmentation. The experimental results show that the optimized SGNN gets better segmentation results and outperforms the existing methods for its distinguished advantages of perfect segmentation without any manual intervention, high self-learning capacity, less computational complexity, robustness to noise, etc. What's more, the experimental results suggest that the proposed method can be widely used in segmentation of all typical images, such as IR (Infrared) images, visible images, X-ray images, and MR (Magnetic Resonance) Images.
Detection of space image target based on improved fractal technique
Balin Tian, Jianping Yuan, Jing Tian, et al.
In the space surveillance applications, recognition of image targets is very important, one of the most difficult problems in image processing is detecting objects in an image. Target detection is a key step for the applications of space image target recognition. An improved method presented in this paper is designed to detect man-made objects in digital images representing in natural environments. This method is based on Fractal theory. According to the fractal feature of man-made object in space, the object can be partitioned off background based on the difference of intrinsic fractal features between them, where the value of fractal dimension (FD) at the objects' edge is much larger than that in other area. The validity of the proposed method is examined by processing and analyzing images of space target at the end of the paper. In detection of targets for real images of Shenzhou-VI and launch vehicle ChangZheng(CZ), the algorithm has a successful outcome. It will provide great efficiency and speed in detection of space image target, which could be also used for further image feature extraction and image segmentation.
Approach to image segmentation with edge gradient estimate
Yue-e Li, Yang Xu, Yingming Chen
Gradient amplitude and phase can't be used together in traditional edge detection procedure, So an algorithm for image segmentation based on edge gradient estimate is given in this paper. First, a standard template is ascertained by the gradient phase of the center pixel; then, estimate edge gradient is the absolute value of the correlation coefficient between the normalized data vector and the template. The method of non-maximum suppression is used when determining pixel local maximum point; hysteresis threshold is used when determining edge point. In both the treatment process, the normalized gradient amplitude and phase would regard as distinguish basis, so this will increase the capability of weak edge detection, and it can suppress the noise impact. This paper gives the relevant experiment result.
Single level set based fast images segmentation model with multiple regions
In traditional image segmentation models based on single level set function, only two regions can be identified because different regions are identified by the signs of single level set function. Though several segmentation models with multiple regions have been proposed, but the largest number of regions that can be identified was limited by the number of embedded level set functions in them. Moreover, the more embedded level set functions, the higher the time cost, usually increasing linearly with the increase of embedded level set functions. In this paper, by introducing the segmentation-measure function, a new model for multi-regions image segmentation based on single level set function is proposed. At the same time, a new initialization function for the level set function is also proposed in order to reduce the time cost of the segmentation model. The experiment results show that the new Image segmentation mode with multiple regions proposed in this paper performs well and dramatically reduces the time cost compared with the popular model for multiple regions proposed by Vese and Chan.
Adaptive recursive algorithm for infrared ship image segmentation based on gray-level histogram analysis
Xinyu Wang, Huosheng Xu, Heng Wang
The infrared ship segmentation in digital images is a fundamental step in the process of ship recognition. This paper presents an adaptive recursive algorithm for infrared ship image segmentation based on the gray-level histogram analysis of the image. The proposed algorithm consists of four phases. First, the gray-level histogram of the image is generated and de-noised by using wavelets transform. Second, a threshold level which best extracts the ship from the water region is selected according to the histogram profile analysis. Third, the rationality of the selected threshold is analyzed based on the prior information about infrared ship images. If the selected threshold is not reasonable, we can still use it as the recursive initial threshold and the infrared ship image will be further segmented with a local recursive method based on the method proposed by OTSU until it reaches the prescriptive termination criteria. Finally, we eliminate the spurious pixels by extracting the greatest connected region and filling the holes. The segmentation algorithm works successfully for classification of infrared ships, and some experimental results are also presented.
Real time peaks extraction
Xiao Zhou, Ruolan Hu, Guilin Zhang
Peaks extraction is a kind of post-process in many image application or vision tasks that can be used for finding the optimum solution in the solution space. In this paper a real time method is proposed. A candidate queue is first build for containing highest peaks in the image in ascending order. Then the image is scanned in sequence. At each scanning position every candidate in the queue is updated respectively by some criterions given in this paper. After the image is scanned over, the highest peaks in the image is achieved in the queue. All the process can be accomplished by logic circuit, so the method is very suitable for hardware system such as FPGA and so on.
Road extraction based on Scansnake from Beijing-1 image
Jianming Gong, Xiaomei Yang, Min Wang, et al.
Beijing-1 small satellite image of 4m high resolution not only makes it possible to extract the detailed information that is difficult to be obtained from low-resolution images, but also brings out new research subjects for automatic extraction and identification of thematic information. The reason for this are as follows:(1) the obvious increase of data requires higher spatial and temporal efficiency of image data retrieval, display, processing, etc.; (2) due to the highly detailed information of high resolution image, under the influence of the Bidirectional Reflectance Distribution Function (BRDF), different parts of the same object may have different grey levels; together with factors such as object shadow, mutual cover, and cloud cover, the phenomenon of same object, different spectrum of high resolution images becomes more prominent, and the different object, same spectrum still exists, which brings greater difficulties to the work of information extraction [1][2]. The road of high resolution image has the following features: (1) the width of the road varies slightly and slowly; (2) the direction of the road varies slowly; (3) the local mean grey level of the road varies slowly; (4) the road is relatively long. Due to the increase of the resolution, the noises such as bridges, cars and trees along the road and its edge also increase. The paper proposes a new road extraction algorithm namely Scansnake aimed at the features of Beijing-1 images. A large amount of experiments proved that Scansnake algorithm has the advantage of object tracking, and under a series of complex conditions such as the variation of the width of the road and the change of grey feature distribution, Scansnake method can extract the road information of the high resolution Beijing-1 image
Robust face tracking algorithm with occlusions
Zhanqing Wang, Youfu Fan, Guilin Zhang, et al.
We propose an adaptive model update mechanism for face tracking based on mean-shift, we employ the Kalman filter to predict a proper original position for mean shift tracking algorithm. To overcome the problem of appearance change, an adaptive modal update is introduced. We classify the occlusion problems into two main cases specified as partial occlusion and complete occlusion according to the number of similar sub blocks between object and candidate. We fuss Kalman predictor into Mean-shift tracker in case of partial occlusion, for case of full occlusion, we divide object and candidate into four parts respectively, according to the previous exact tracking result, we compute the average velocity of the target, and then check the condition for face reappearing, with which we present an efficient target search strategy to deal with full occlusion. Various tracking sequences demonstrate the superior behavior of our tracker and its robustness to appearance changes and occlusions.
Multiscale kernel method for image matching
Hui Cheng, Jing Zhou, Shian Ma, et al.
In this paper, a new multiscale kernel function model is proposed, according to kernel function nature and the wavelet frame theory. Thus a new image support feature is defined based on the multiscale wavelet kernel regression for the purpose of improving image matching algorithm. The comprehensive the feature match and grey correlation method, a novel image fast matching algorithm is proposed based on support feature points. The partial Hausdorff distance is adopted as a similarity measure combined with the wavelet decomposition iteration to search the fine strategy, in the wavelet domain. According to support feature points the position in original image, extracts corresponding the grey value, and achieves the correlation matching. The matched data are compressed effectively because support feature points are sparse. Experimental results demonstrate that the proposed algorithm is robust, fast and can achieve matching accurately.
Moving point target detection using temporal variance filter in IR imagery
A crucial problem in Infrared Search and Track (IRST) systems is the detection of moving point targets, there are many algorithms reported in the literature for dealing with this problem, yet none of them yield acceptable results under all situations. In this paper, we describe a new temporal variance filter (TVF) for detecting targets whose velocity are higher than 1 pixel/frame; the filter iteratively estimates the temporal variance of each pixel, then subtracts the last iteration step variance from the variance of current step. Subsequently, we introduce a novel image segmentation algorithm in order to extract point targets from clutter background, the trajectories of the point targets could be established by post-processing algorithm. Before applying the temporal filter, the anti max-median filter given by Suyog D. Deshpande et al. is incorporated as a preprocessing technique to suppress cloud clutter. It is assumed that targets' velocity is higher than 1 pixel/frame; targets with sub-pixel/frame velocity are not considered. The performance of our approach is evaluated by using available real-world infrared image sequences containing simulated moving point targets; it performs steadily under most situations.
Rapid and effective method of quality assessment on sequence iris image
Yi Huang, Zheng Ma, Mei Xie
This paper proposes a multiple-step method of quality assessment on sequence iris images. Based on the spatial domain and frequency domain of image feature-vector space, the method sets different criteria in different steps to eliminate poor-quality images: firstly to judge the intensity and clarity of iris images roughly; secondly to preprocess iris images by making morphological operation, extract the region of interest(ROI) which covers pupil by blocking image and then make binary-conversion of iris image and locate pupil; thirdly to select different ROI, analyze the factors affecting image quality, such as defocus blur, motion blur, eyelid closure, eyelashes shelter, dynamic characteristics of eyeball, etc., and eliminate substandard images from the target sequence; lastly to set a comprehensive criteria to select the best quality image from the sequence. The method is verified by 300 sequences of iris images and experimental results show that only through locating pupil instead of utterly locating the outer edge of iris can this method quickly and precisely judge whether iris is sheltered and blurred or not, and that each of its steps can almost eliminate substandard images, thus not only reducing the number of iris images to be assessed but also saving the time of processing images.
Automatic recognition of earthquake-caused building damage in cities using multispectral image fusion
Haihui Wang, Hui Du, Minjiang Chen, et al.
Automatic recognition based on image fusion techniques are widely used to integrate a lower spatial resolution multispectral image with a higher spatial resolution panchromatic image. The earthquake events were first researched after the Kocaeli Earthquake of 1999 show that the spatial images from various satellites could be exploited. The remote sensing which in terms of spatial resolution and data processing open new possibilities concerning the natural hazard assessment. However, the existing techniques either cannot avoid distorting the image spectral properties or involve complicated and time-consuming frequency decomposition and re-construction processing. To address these problems, we present our study on a HIS transform and intensity modulation algorithm. The algorithm is further optimized a proposed objective of minimizing error rate. Experiments in recognition of building damage due to earthquakes applications show that the algorithm provides better recognition accuracy than others. Although some environment problems, such as the influence of sunshine need further research, the proposed method can benefit further study of the application.
Simulated ship recognition using two-dimensional PCA
Guangzhou Zhao, Guangxi Zhu, Feng Peng, et al.
This paper proposes a fast and robust algorithm for classification and recognition of ships based on the two-dimensional Principal Component Analysis (2DPCA) method. The three-dimensional ship models achieve by modeling software of MultiGen, and then they are projected by Vega simulating software for two-dimensional ship silhouettes. The 2DPCA method as against conventional PCA method for simulated ship recognition using training and testing experiments, as the training and testing sample size is large, and there are great variations in different azimuth and elevation for ship viewpoints. The experiment of ship recognition using the global feature of ships is not satisfied with us, so we proposed an improved 2DPCA method based on the local feature of ships. Some recognition results from simulated data are presented, it shows that the improved 2DPCA method outperform PCA in ship recognition and also superior to PCA in terms of computational efficiency for feature extraction. So our method is more preferable for ship classification and recognition.
Real time tracking by LOPF algorithm with mixture model
Bo Meng, Ming Zhu, Guangliang Han, et al.
A new particle filter-the Local Optimum Particle Filter (LOPF) algorithm is presented for tracking object accurately and steadily in visual sequences in real time which is a challenge task in computer vision field. In order to using the particles efficiently, we first use Sobel algorithm to extract the profile of the object. Then, we employ a new Local Optimum algorithm to auto-initialize some certain number of particles from these edge points as centre of the particles. The main advantage we do this in stead of selecting particles randomly in conventional particle filter is that we can pay more attentions on these more important optimum candidates and reduce the unnecessary calculation on those negligible ones, in addition we can overcome the conventional degeneracy phenomenon in a way and decrease the computational costs. Otherwise, the threshold is a key factor that affecting the results very much. So here we adapt an adaptive threshold choosing method to get the optimal Sobel result. The dissimilarities between the target model and the target candidates are expressed by a metric derived from the Bhattacharyya coefficient. Here, we use both the counter cue to select the particles and the color cur to describe the targets as the mixture target model. The effectiveness of our scheme is demonstrated by real visual tracking experiments. Results from simulations and experiments with real video data show the improved performance of the proposed algorithm when compared with that of the standard particle filter. The superior performance is evident when the target encountering the occlusion in real video where the standard particle filter usually fails.
Unsupervised texture segmentation based on nonsubsampled contourlet and a novel artificial immune network
Wenlong Huang, Licheng Jiao
This paper describes a novel structural adaptation artificial immune network (SAAN) clustering algorithm for texture segmentation. In the SAAN, a new immune antibody neighborhood and an adaptive learning coefficient are presented. The model can adaptively map input data into the antibody output space, which has a better adaptive net structure. Images are first partitioned into a set of regions by using the watershed segmentation. Then the nonsubsampled contourlet texture features are extracted from each watershed region as the antigens of the SAAN. Finally the antibodies clustering results of the SAAN are combined to yield a global clustering solution by the minimal spanning tree, which need not a predefined number of clustering. The experimental results with various texture images illustrate the effectiveness of the proposed novel segmentation algorithm.
Motion estimation and geometric active contours for object tracking
We present a robust geometric active contour model to track targets in video sequences captured from mobile cameras. The target's contour is tracked on each frame of the sequence by regional information. The regional information is calculated by color histogram. The matching criterion is formulated by minimizing the Bhattacharyya coefficient between the color histogram of reference target and that of the background. The contour's evolution is implemented using the geometric active contour algorithm and the level set method. For each frame coming from the sequence, motion estimation is done before the contour evolution process. To make so, Kalman prediction and template matching are performed as the motion estimation technique to locate the region of interest (ROI). The robustness and effectiveness of the proposed algorithm is demonstrated on real sequences.
Application of a registration method based on SVD in detecting moving object of dynamic background
This paper proposes a method used to detect big moving object in the complicated dynamic background, which integrates the phase correlation technique including singular value decomposition and the method in which multi-frames difference images is multiplied. The phase correlation algorithm based on singular value decomposition is insensitive to noise and change of gray and contrast. Comparing with many complex phase correlation algorithm and registration algorithm in spatial domain, our method not only can effectively restrain noise, but also enhancing the registration precision, whose speed is nearly two times as quickly as original phase correlation algorithm. The fact is found by the result of experiment that the phase correlation matrix is rank one for a noise-free rigid translation model. A new phase correlation matrix is recast based on the property which can effectively restrain noise and change of gray. By estimating global moving vector of two images using phase correlation based on singular value decomposition, background is accurately matched. The matched images are processed to calculate the image differences between the first and fourth, the second and fifth, the third and sixth. After these difference images are multiplied, clear edge of moving object is obtained. Thus the accurate location of object is realized by calculating barycentre of image. At last, simulation results prove that this proposed method can overcome effectiveness well in the lighting variations and noise. It is also efficient and applicable for accurate moving object orientation in the complicated dynamic background.
Research on benthic scene recognition using multi-scale self-similarity model and statistical analysis of increments
Guoliang Yang, Fuyuan Peng, Xutao Li, et al.
In this paper, we analyzed the increment distribution and the self-similarity behavior of texture images of three kinds of particular underwater objects during the mineral hunting process. The experimental data has shown that the H exponent of real underwater natural texture is not a constant over all scale range, but a variable with respect to the measure scale or time index. In order to investigate the multi-scale self-similarity behavior of the objects, we had extended the traditional FBM so that the self-similarity parameter H is taken as a variable H(s) with respect to measure scale s. The class-separability of self-similarity feature is measured, and the feature selection criterion is given. Pattern classification simulation experimental results have shown the effectiveness of the selected feature set combining the self-similarity parameter HΔ(3), the variance D(HΔ) and the increment variance AD. The correct ratio is up to 96% on average, which can be used in automatic detection and recognition for AUVs to complete their tasks.
Data mining application based on fuzzy clustering for traffic network evaluation
In GIS for Transportation (GIS-T), how to discover knowledge from complex traffic data is very vital. This paper focuses on the regional traffic to evaluate the traffic condition in a certain region, which can provide decision-making support for leadership. Nowadays, there are multitudinous regional traffic network evaluation models, most of which are based on a single item of index. It is difficult to give a satisfying evaluation result to the whole regional traffic condition. In this paper, we establish a regional traffic evaluation system for traffic network based on the theory of fuzzy clustering and maximizing deviation, and evaluate the traffic networks of 15 regions in Hubei province in 2004.
Adaptive image fusion based on nonsubsampled contourlet transform
Multiresolution-based image fusion has been the focus of considerable research attention in recent years with a number of algorithms proposed. In most of the algorithms, however, the parameter configuration is usually based on experience. This paper proposes an adaptive image fusion algorithm based on the nonsubsampled contourlet transform (NSCT), which realizes automatic parameter adjustment and gets rid of the adverse effect caused by artificial factors. The algorithm incorporates the quality metric of structural similarity (SSIM) into the NSCT fusion framework. The SSIM value is calculated to assess the fused image quality, and then it is fed back to the fusion algorithm to achieve a better fusion by directing parameters (level of decomposition and flag of decomposition direction) adjustment. Based on the cross entropy, the local cross entropy (LCE) is constructed and used to determine an optimal choice of information source for the fused coefficients at each scale and direction. Experimental results show that the proposed method achieves the best fusion compared to three other methods judged on both the objective metrics and visual inspection and exhibits robust against varying noises.
Human face detection and tracking based on supervised learning
Min Luo, Xiaohui Duan, Shiwen Zhu, et al.
In this paper a novel method of human face detection and tracking based on supervised learning for video sequence is designed. The system is composed of a face detector using boosted rectangular filters with a new representative based integration method, a linear capture model and a quadric tracking model. The main contribution of this paper is a new view to face tracking solutions on condition that a robust real-time detector is adopted first. It differs fundamentally from traditional tracking algorithms for that it organically combines fast and robust detection with efficient capture and tracking which can be easily implemented in practical video systems while obtaining a satisfying real-time performance. Experimental results show that this algorithm can finely meet the reliability and effectiveness demands of video surveillance system.
Research of robot simultaneous localization and mapping in multiple mobile robot system
Yanbiao Huang, Yimin Yang, Qicheng He, et al.
It needs an exact global map when the multiple mobile robot system making decisions in the task allocation and action control, but each robot can only obtain the information which surrounds him nearby and often lose the information at long bowls, in other words, the robot can not build a comprehensive global map of whole field. Aimed at these problems, a multi-sensor data fusion subsystem was designed and added into the multiple mobile robot system. The experiment shows the whole system's fault-tolerance capability and identifying capability are both enhanced evidently.
Local mapping for the middle-size league of RoboCup
Qicheng He, Yimin Yang, Xuexi Zhang, et al.
Aiming at making up the limitation of catadioptric vision system that lacks of nearby image information, which is used for the middle-size league of RoboCup - The World Cup of Soccer Robots, this paper introduces the monocular front vision process with the purpose of improving the primary system. The paper puts forward the idea of the monocular front vision process at first, explains the algorithm of every module in detail, and comes up with the result as well as analyzes it according to the experiment at last.
Fast matching method between infrared image and optical image
Jing Liu, Jiyin Sun, Junlin Zhu, et al.
In this paper, we present a high performance, coarse-to-fine scheme for image matching of infrared image and optical image. In order to eliminate the gray level difference and tolerance to distortion and noise existing in practice, this scheme uses the edge feature and combine a new similarity measures with modified Hausdorff distance to achieve the coarse-to-fine matching scheme. Our proposed method firstly extracts the feature points based on the method of Monte Carlo that reduce the computation load for the next matching, and then a new similarity measure is defined for coarse matching. Based on existing method of Monte Carlo evaluation the Huasdorff distance (MCHD), we define Monte Carlo modified Hausdorff distance (MCM-HD) to achieves the fine matching. Experiment are performed on a large test set and the result show that the fast search method diminishes the number of positions for calculate of Hausdorff distance, thus the computational load reduces and it is helpful to use Hausdorff distance in real time image matching. Compare with MCHD algorithm, our proposed method effectively improves the precision and reduces the execution time.
Region growing based road extraction in SAR images
Yuan Guo, Zhengyao Bai, Yue Li, et al.
A new region growing based algorithm for extracting roads in spaceborne synthetic aperture radar (SAR) images is presented in this paper. The basic idea is that an initial seed point moves in a determined direction with a certain step length. The algorithm is composed of five sequential steps: the first is to search initial seed point and the second is to choose a circular window size similar to the width of the first road. In the third step, the criteria for region growing are chosen, and the fourth is to determine the key elements of new roads growing. At last, stopping rules are specified. The experimental results show that the proposed method can better extracted roads in SAR images.
Forward-looking recognition based on convex hull invariants of oil depot region
Fangfang He, Jiyin Sun, Bing Han, et al.
Forward-looking navigation system is a fire-new technique for terminal guidance of intending precision-guided weapons and research on oil depot recognition of forward-looking imaging is an essential task for this control and guide system. As conventional matching methods could not overcome perspective transmutation, a new method to identify the forward-looking area of oil depot was advanced in this paper. First, constructed three statistics of regions based on convex hull, which were invariant to affine transform. Then, number of inside oilcans could easily be achieved by adding a decision step. Finally the area of oil depot could be located according to the comparison between the computed number and the foreknowable number under a given threshold. Experiments applied to optical images in different areas show that the proposed method is accurate and has wider application in identifying such small objects as oilcans, and it realizes automatically recognizing area of oil depot from forward-looking imaging.
Road extraction from SAR images based on particle filtering
Road extraction plays an important role in many applications such as traffic monitoring. In order to speed the extraction and enhance its precision, an approach based on particle filtering is proposed in this paper. Firstly, an improved line detector is presented to extract road candidates, which makes use of the road characteristic on SAR images. Then particle filtering based on Monte Carlo theory is applied to group the candidates. Applied results show that the road extraction method is effective and the road features on SAR images have been extracted accurately. Moreover, the method can be realized simply and save the amount of calculation.
Adaptive four windows wavelet image denoising based on local polynomial approximation-intersection of confidence intervals
Hongxiao Feng, Biao Hou, Licheng Jiao, et al.
Local Polynomial Approximation-Intersection of Confidence Intervals (LPA-ICI) is a new approach, which can find the boundary of the isotropic region efficiently, especially for noisy images. This paper presents a novel image denoising method, adaptive four windows wavelet image denoising based on LPA-ICI, which is composed of three parts: searching for four adaptive windows with LPA-ICI, updating the noisy wavelet coefficients by hard threshold and obtaining a final "clean" pixel value by fusing the updated pixels with different weights which are determined by the sparsity of regions. Experiments show that our algorithm has advanced performance, reconstructed edges are clean, and especially without unpleasant ringing artifacts.
Interactive dynamic graph cut based image segmentation with shape priors
Chen Liu, Fengxia Li, Shouyi Zhan
In this paper, we present a new method bases on dynamic graph cut and captures both the shape and the nature information of the image for interactive image segmentation. While traditional interactive graph cut approaches for image segmentation are often successful, they may fail in camouflage. Prior shape knowledge can largely mitigate this problem. In this paper, two kinds of shape priors are taken into account to obtain more accurate results. In order to use the information from user input more effectively, a weight function is introduced to control the relative importance of shape knowledge. Then, a one-shot fully dynamic graph cut algorithm is introduced to minimize the energy function, and during this procedure, only a subset of pixels in the image is considered, which greatly reduces the complexity of dynamic graph cut algorithm. Extensive experiments, including comparisons with some state-of-the-arts, show the effectiveness of our methods in improving the segmentation performance and saving the processing time.
SAR image ATR using SVM with a low dimensional combined feature
Hongqiao Wang, Fuchun Sun, Zongtao Zhao, et al.
In this paper, a grayscale wavelet moment and entropy combined feature, which can well represent the images with much lower dimensions, is proposed for the MSTAR targets' classification, also a discriminative feature selection and evaluation method for multi-class targets is presented. By introducing SVM as the classifier to some simulation tests, it can be shown that the grayscale wavelet moment and entropy combined feature has good capability for translation, scaling and rotation transformation in both local information and global information conditions, by the reasons of wavelet moments' multi-resolution analysis, moment invariant quality and entropy's statistic quality for image disorder. The test also confirms that this feature has a significant improvement on classificatory accuracy using SVM with a lower feature dimension than the other features such as the single wavelet moment, Hu's moment and PCA.
Fast alignment of marks in COG bonding based on multi-resolution
COG (chip on glass) bonding is widely used in LCD industry. The alignment of marks in COG bonding needs high precision and reliability. It is of utmost importance to find two marks in COG fast and exactly. Its key technique is to use the advanced optical system to get the two-marked positions of chip on the glass and to use the PLC procedure control to complete the automatic alignment A series of experiments are performed to test the algorithms proposed, which show that the fast alignment of marks in COG bonding based on multi-resolution is both rational and highly effective.
Recognition algorithm for characters at ends of steel billet using features of character structures
For the requirement of identification and trace of steel billet in the procedure of manufacture and management, the ends of steel billets are all printed a series of characters. So, each end of steel billet must be tracked and recognized automatically before they are imported to furnace. Owing to the effect of the bad-imaging environment, there are a lot of difficulties in recognizing characters at ends of steel billet. So, the real-time automatic recognition of characters at ends of steel billet is different from the recognition of the license plates. For the former methods dissatisfied with practical demands, an identification method using features of character structures is introduced by this paper, which greatly lowers error rate and rejection rate effectively.
Infrared small target detection and tracking under the conditions of dim target intensity and clutter background
Xiangzhi Bai, Fugen Zhou, Ting Jin, et al.
To reduce the influences of the dim target intensity and heavy clutter on infrared small target detection and tracking, a novel algorithm is presented in this paper. The algorithm proposes a modified top-hat transformation by importing the property of the small target region firstly, which largely enhances the dim target and apparently suppresses the heavy clutter. Consequently, the potential targets are easy to be segmented by the iterative thresholding method. After decreasing the false alarms through the dilation cumulation, the real target and the trajectory are correctly given by using the data association formed by the motion property of the real target. Various experiments verified that the proposed algorithm was efficient and robust for dim target detection and tracking under the condition of heavy clutter.
The abnormal behavior analysis of single person on the road based on region and behavior features
Wei Wang, Runsheng Wang, Yiwen Chen
In this paper, a method to detect whether the behavior of a single person in video sequence is abnormal is proposed. Firstly, after the pre-processing, the background model is gotten based on the Mixture Gaussian Model(GMM), at the same time the shadow is eliminated; then use the color-shape information and the Random Hough Transform (RHT) to abstract the zebra crossing and segment the background; Lastly, we use the rectangle and the centroid to judge whether the person's behavior is abnormal.
Detection of ash fusion temperatures based on the image processing
Peisheng Li, Yanan Yue, Yi Hu, et al.
The detection of ash fusion temperatures is important in the research of coal characteristics. The prevalent method is to build up ash cone with some dimension and detect the characteristic temperatures according to the morphological change. However, conditional detection work is not accurate and brings high intensity of labor as a result of both visualization and real-time observation. According to the insufficiency of conventional method, a new method to determine ash fusion temperatures with image processing techniques is introduced in this paper. Seven techniques (image cutting, image sharpening, edge picking, open operation, dilate operation, close operation, geometrical property extraction) are used in image processing program. The processing results show that image sharpening can intensify the outline of ash cone; Prewitt operator may extract the edge well among many operators; mathematical morphology of image can filter noise effectively while filling up the crack brought by filtration, which is useful for further disposal; characteristic temperatures of ash fusion temperatures can be measured by depth-to-width ratio. Ash fusion temperatures derived from this method match normal values well, which proves that this method is feasible in detection of ash fusion temperatures.
Analysis of wavelet image denoising model in Besov spaces
Qibin Fan, Minkai Jiang, Wenping Xiao
In this paper, we discuss the image denoising model which DeVore et al. had established, in which both distance and smoothness can be measured by the objective function, and analysis the model for wavelet image denoising in the Besov spaces with p = q. In addition, we give the exact thresholds for the model, and prove that for 0 < p <1 the effect of noise removal using our methods is in between hard wavelet shrinkage and soft wavelet shrinkage. For the case 0 < p < 1 and 1 ≤ p ≤ ∞, which refers to the problems on the convergence of the iteration of the equations and on the complexity of computation, we give the simplified algorithms. Comparing the threshold given by this paper with Lorenz threshold, we conclude that the former is more meticulous than the latter for the model.
Local level set segmentation method combined with narrow band
Yushi Li, Jun Zhou, Junlong Li, et al.
The paper deduces a general form of energy function from the level set method based on Mumford-Shah model. It introduces the gradient of image features into the energy function, which could make the segmentation more precise and the algorithm converge faster. In order to detect local object in the image with clutter background, a local level set segmentation method using the proposed energy function is presented in this paper. The method combined with narrow band could obtain local optimal segmentation, which just needs prior location of the object. To tackle the problem that the calculation cost of level set is so expensive, the paper proposes an efficient algorithm for narrow band which implements very fast. The algorithm starts with a simple initial curve, and then it only updates the level set function in the narrow band. The local level set method is applied successfully to image segmentation with cluttered background, multi-object detection and moving object detection. The results of the experiments are presented in the end of paper.
Automatic structure detection in a point-cloud of buildings obtained by terrestrial laser scanning
Qingming Zhan, Qiancong Pang, Wenzhong Shi
In recent years, terrestrial laser scanner (TLS) has become a popular data acquisition tool for producing irregularly-spaced point clouds as well as airborne laser scanning (ALS). Automated detection of structures (roof and ground etc.) based on the point cloud analysis of buildings has become increasingly important. One of the most demanding tasks in TLS is the filtering of the ground and roofs in TLS point clouds. This paper proposes a method for detecting buildings' structures from an irregularly-spaced point-cloud. This method is consisted of segmentation and classification. As the previously developed the segmentation methods can not be applied to it directly, it has to perform twice pre-filtration so as to proceed to further calculation for segmentation and classification. More importantly the algorithm is extensible and future work will further strengthen the algorithm.
3-band symmetric bi-orthogonal filter banks with compact support and their applications in signal processing
Guokuan Li, Changfa Xu, Jun Guan, et al.
Multi-band wavelets are newly emerging branch in wavelet family and could have better properties than dyadic wavelets in terms of symmetry, orthogonality, compact support and smoothness. The purpose of this paper is to present a new method for constructing the filter banks of 3-band symmetric bi-orthogonal wavelet using a scaling function of linear spline function. To construct such 3-band wavelet with desirable properties, a set of linear algebra equations can be listed according to the requirements of the bi-orthogonal multi-resolution analysis. And these equations are then solved to obtain the filter coefficients. The properties of the filters and the multi-resolution analysis (MRA) in signal processing are discussed. Experiments show that the 3-band filter banks could be potentially better in signal processing than dyadic wavelets.
Fast algorithm for document skew detection method using run-length smoothing, Hough transform, mathematical morphology and wavelet
Fu-sheng Guo, Xiang-ning Chen, Jiangbo Li
Document skew is often introduced into the scanned or printed document, so the document skew detection and correction play an important role in the document image processing. This paper gives a whole review of existent algorithms and describes their respective shortcomings. We propose a novel algorithm combined with run-length smoothing, Hough transform, Mathematical Morphology and Wavelet, as to improve its precision and speed. The algorithm this paper gives is feasible and effective, and it can be run automatically. It can be faster and more accurate than other methods and can be applied in document words identify.
Multi-focus image fusion using adaptive Wiener filter
This paper presents a new method for multi-focus image fusion. In the method, the source images are first decomposed into blocks, and the decomposed images are then combined by the use of adaptive Wiener filter. Effects of the block size and threshold are analyzed, and comparison with wavelet transform based method is done. Experimental results show that the proposed method is comparative to wavelet transform based methods for the images without noise, while this method is computationally simpler, and can be implemented in real-time applications. Experimental results also show that under noise circumstances, additive noise or multiplicative noise, the proposed method is obviously superior to the wavelet based method.
Aerial sequence image mosaic using reduced sift descriptors
Hongya Tuo, Zhongliang Jing, Tinghou Zhang
Sequence image mosaic is an important and effective method to build a large "panoramic" scene which includes two main steps: image registration and intensity blending. In this paper, SIFT feature points are used to match images. SIFT are invariant to rotation, translation and scale changes, but the significant drawback is the high dimensional feature descriptor which lead to the expensive computation. So reduced SIFT descriptors are proposed to increase the speed of image registration. Linear combination methods of the matching points' gray-values are used for intensity blending. The experiments show that our method is useful and has high registration accuracy.
Fingerprint segmentation based on local Fourier transform
The segmentation of fingerprint images plays an important role in fingerprint recognition. A new algorithm based on Local Fourier Transform (LFT) for the fingerprint segmentation is proposed in this paper. Firstly, we perform the Local Fourier Transform on image to get eight independent Local Fourier coefficients per pixel. Then, block features are extracted by calculating the 2nd, 4th, 6th order moments of the local Fourier coefficients of every pixel in the block. After that, a Fisher linear discriminant classifier is trained for the classification per block. Finally, mathematical morphology and region boundary smoothing is applied as postprocessing to obtain compact clusters and to reduce the number of classification errors. The experimental results on the databases of FVC2004 demonstrate the robustness and the efficiency of the proposed method.
Template coarse matching of SAR and optical images based on wavelet subbands and Hausdorff distance
Xia Mao, Kang Huang, Yaoming Liang
This paper presents a suited approach to deal with the template coarse matching of synthetic aperture radar (SAR) and optical images based on wavelet subbands and Hausdorff distance. Firstly, we analyze the discrepancy of imaging character between SAR and optical images. Secondly, we occupy the max flat wavelet decomposition to obtain the template images, which also can suppress the speckle in SAR. Thirdly, improved canny algorithm is employed to extract the edge feature of SAR and optical images after image enhancement, respectively. Finally, this paper applies the modified Hausdorff distance with the optimum coefficient to the similarity measurement for the matching. The experimental results demonstrate the validity of the proposed methodology.
Non-rigid object tracking using adaptive part-based model
Xiaohui Shen, Jin Zhang, Jie Zhou, et al.
This paper presents an adaptive part-based probabilistic model for non-rigid object tracking. Without any assumption on scenes or poses, our model is online generated and updated. The parts in the model are extracted by clustering based on the appearance consistency of local feature descriptors in the object. A probability indicating the possibility of a part belonging to the object is then assigned to each part and adapted during tracking. We also propose a fully automatic algorithm for single object tracking with model matching and adaption. Our approach is evaluated on three different datasets and compared with previous work on visual tracking. The experimental results showed that our approach can track non-rigid object under occlusion and object deformation effectively in real time. Moreover, it works even if the target is partially occluded at initialization step.
Research on classification of high-resolution remote sensing image
Xiwang Zhang, Bingfang Wu, Qiangzi Li
High-resolution remote sensing image is usually considered that its pixel size is less than 10 meters. Traditional classification methods based on pixels are not fit for the classification of this kind of image because this kind of image has higher spatial resolution and more local heterogeneity compared to the low-resolution remote sensing image data. Object-oriented image classification method provides a good technique to solve this problem. This method segments image to create homogeneous regions or image objects through region merging or boundary detection algorithms. Objects possess more features such as geometric and structure characteristics besides spectral characteristics than pixels. So it is important to select appropriate characterstics in classification. Class-Related features, landscape pattern metrics, geometric attributes of objects, spatial information are very useful characteristics. The paper will pay more attention to the selection and integrative utilization of these features and spectral characteristics, and give several examples to show their performance. (1) If We want to extract a kind of feature which has similar spectral characteristic as the other feature but has a certain positional relationship with a specific feature, at this time, Class-Related features will be very efficacious. (2) Both river and pounds have also similar spectral characteristic, but has different geometric characteristic in the landscape pattern metrics. Synthetically use of the landscape pattern metrics and spectral characteristic will wok well. (3) In the same segmentation scale, the objects from the region with more homogeneity will be bigger than other objects from the region with more heterogeneity. So, the area and spectral characteristic can be used in classification. The results show a better accuracy. The selection and integrative utilization of features of objects were very important in achieving these high accuracies.
Image registration method based on evolutionary modeling with constraints
Zongyue Wang, Hongchao Ma, Jianwei Zhang, et al.
Considering the deficiency of mapping model in traditional image registration, a new image registration method based on evolutionary modeling is proposed in this paper. Multi Expression Programming has been used as modeling tool to build mapping model. To avoid over fitting and improve actual effective, constraints of the mapping function's slope and curvature were added during modeling process. SAR image and optical image rectifying experiment is given in the last. The experiment result indicated that the evolutionary model has high precision for image registration. This method is fit for image registration.
Visual tracking by threshold and scale-based particle filter
Hui Yin, Yongfeng Cao, Hong Sun, et al.
Particle filter has attracted much attention due to its robust tracking performance in clutter. However, a price to pay for its robustness is the computational cost. Meanwhile there is no exact mechanism for choosing or updating scale in its framework for accurate tracking. In this paper we propose a threshold and scale based particle filter (TSPF). It employs a threshold to discard the bad particles and keep the good ones. In this case, the efficiency of particles is improved and the number of required particles is greatly reduced. It also adapts Robert T. Collins's theory of selecting kernel scale for mean shift blob tracking to particle filter. Experiments show TSPF works well, both spatially and in scale.
New color management model for digital camera based on immune genetic algorithm and neural network
Color management for scanner is one of the key techniques in the color reproduction in information optics. A new digital Camera color management model is presented based on analyzing its rendering principle. First, standard color target is taken for experimental sample and substitutes color blocks in color shade district for complete color space to solve the difficulties of experimental color blocks selecting. Second, Immune Genetic Algorithm is presented to correct BP neural network to speed up the convergence of the model. Finally the experimental results show that the model can improve scanner color management accuracy and can be used in digital Camera color management practically.
High resolution satellite imagery segmentation based on adaptively integrated multiple features
Aiping Wang, Peng Tian, Shugen Wang
Automatic segmentation of high resolution satellite (HRS) imagery is the first step and a very important part of object-oriented approaches. The HRS sensors increase the spectral within-field heterogeneity and the structural or spatial details of images. Spatial features are important to HRS image analysis in addition to spectral information. This paper presents a novel feature extraction method and evaluates its performance on segmentation of HRS images based on adaptively integrating multiple features. The first two principal component (PC) images are obtained by principal component analysis (PCA) of a multispectral image and used to calculate the texture and spectral distributions of a region, which are denoted by two-dimensional (2D) histograms. The 2D texture histogram of a region is the joint distribution of its two texture labeled images calculated by rotation invariant local binary pattern (LBP) operator. The spectral distribution of a region is the joint distribution of the pixel values of its two PC images after normalization. The color feature is a 2D hue/saturation histogram that is computed through IHLS color space. The three features are integrated by a weighted sum similarity measure and used to hierarchical splitting, modified agglomerative merging and boundary refinement segmentation framework. The segmentation scheme based on adaptively integrating multiple features demonstrates promising results.
Multi-scale representations of the motion trajectory
Hong Shu, Jun Pang, Cuihong Qi
At present, mobile computation is developing rapidly for location-based services. For the online GIS or cartography of moving objects, the techniques of progressively refined details greatly reduce the overhead of storage, computation, display, and communication resources. Again, human cognition about the reality is made at multi-scales of abstraction. Multi-scale representations of the moving object trajectory are required upon human hierarchical cognition and progressively refined details modeling. To this end, wavelet-based multi-scale representations of the motion trajectory are presented in this paper. We consider parametric motion and a sequence of spatio-temporal coordinates as database representations of the motion trajectory. Correspondingly, multi-scale wavelets representations are considered as computing representations of the motion trajectory. The wavelet transform is applied to time-series of spatial coordinates for finding dramatic or gradual changes of motion speed at each scale. Multi-scale wavelets representations of the motion trajectory have revealed global motion trends at the large scale and local motion details at the small scale. Inherently, spatio-temporal coordinates, motion, and dynamics are three-scale representations of spatio-temporal semantics. Multi-scale representations of motion trajectory are experimentally illustrated. Our work is devoted not only to mobile computation but also to extension of wavelet analysis into geometry data processing.
Combining color and texture features for image retrieval
Guiting Wang, Baobao Tian, Licheng Jiao
In this paper, a new method named BQCGW (block-based method of combining quantized colors and Gabor wavelet features) is proposed for image retrieval. HSV color space, in which measured color differences are proportional to the human perception of such differences, is quantized into 67 kinds of representative colors. We also propose the use of Gabor wavelet features for texture analysis. Images in the database are divided into nine blocks before extracting color and texture features. Experiment results show that our method is feasible and valid.
Efficient and robust global motion estimation for automatic target recognition
Ming Jin, Bin Xue, Dingming Peng
In the automatic target recognition with complex background, the method of detecting motion target using the global motion estimation (GME) in image sequences is often proposed. Due to the possible presences of differently moving foreground objects and other sources of distortions, improvement of robustness and preciseness of GME is very difficult. Therefore the method of GME based on the M-estimators' formulation with direct multi-resolution is proposed in our paper. The M-estimators' formulation is not only executed in gradient descent at each level of the pyramid, but also applied in initial translation estimation with minimal SSD at the coarsest level, which assures the convergence of the subsequent gradient descent algorithm. Comparative experiments are performed to validate the performance of the proposed algorithm. The effectiveness and improvements can be observed from the comparisons.
Novel spherical panorama creating algorithm based on curve surface mosaic
Xiaohui Li, Yinqing Zhou, Zulin Wang
This paper presents a novel algorithm based on curve surface mosaic, to create a full view spherical panorama from image sequences. The work is concentrated on sphere projection, blank holes elimination, global illumination alignment and curve patch stitching. When in projection, a special longitude-latitude-curve-patch is proposed to describe the projective image to avoid info losing and wrinkle unwrapping that occurred in some traditional methods. Then a way of "inverse-interpolation" is applied to eliminate projective blank holes caused by discrete calculation. To achieve global illumination alignment for patches with great illumination differences, a novel method of "dispersing cumulative error" is presented. It overcomes the shortcoming of traditional ways that are only for neighboring illumination alignment. The final stitching of curve patches is accomplished by using a matching method based on image feature, and a smooth seamless spherical panorama is gained. The whole algorithm runs automatically, which has high performance in illumination alignment and spherical mosaic. It is valuable in practical application.
Efficient motion segmentation for H.264 compressed video
Yu Lu, Zhaoyang Zhang, Zhi Liu, et al.
The H.264 standard is a new state-of-the-art video coding standard with extensive applications. This paper presents a simple and efficient approach for motion segmentation in H.264 compressed video. Several preprocessing steps are used before actual motion segmentation. The raw motion vector (MV) field extracted from H.264 video is first spatially normalized and then accumulated by the forward projection scheme to obtain the dense MV field. The following global motion compensation is performed on the accumulated MV field to acquire the residual MV field. Based on the residual MV field, a hybrid scheme including edge detection and region growing for motion segmentation is proposed. The edge map is used as a mask to guide region growing, which is created by Canny operator based on the magnitude map of residual MV field. At last, hypothesis testing as the major postprocessing technique is exploited to distinguish between the background and different moving objects. Experiment results demonstrate that the high-efficiency performance and good segmentation quality of the proposed approach.
Group tracking using mathematical morphology and multiple clustering hypotheses
Yu Li, Liu Yu
Group tracking is to track groups of objects engaging in a common activity (formations), which is a challenging and significant problem for intelligent visual surveillance. It can develop an overall understanding in some applications. Also, information related to group behavior can be used to enhance the precision and reduce computational complexity for the estimate of individual target. In this paper a group initialization and tracking method is presented. Two stages are involved. First, groups are initialized by clustering detections of targets based on a mathematical morphological method. Second, the groups are tracked in terms of prediction, hypotheses, confirmation and deletion. Satisfactory results are shown when testing the algorithm on simulative sequences.
Automatic image registration based on convexity model and full-scale image segmentation
Kaimin Sun, Haigang Sui, Yan Chen
Image registration plays a critically important role in many practical problems in diverse fields. A new object-oriented image matching algorithm is presented based on the convexity model (CM) and full-scale image segmentation. The core idea of this matching algorithm is to use image objects as matching unit rather than points or lines. This algorithm firstly converts images into image objects trees by full-scale segmentation and convexity model restriction. Because image objects which accord with the convexity model have rich and reliable statistical information and stable shapes, more characteristics can be used in object-based image matching than pixel-based image matching. Initial experiments show that matching algorithm proposed in this paper is not sensitive to rotation and resolution distortion, which can accomplish the image matching and registration automatically.
Intensity-based correlation for heterogeneous images scene matching
Sheng Zhong, Huimin Cao, Tianxu Zhang
Downward looking scene matching is an important technique of the aircraft automation guidance. To solve the heterogeneous images scene matching problem, we present two effective methods based on intensity-based correlation in this paper. One is to search the real match position based on the feature of the peak on the correlation surface. We propose a criterion to search the proper matching. The other is to use a non-linear filter to pre-process the images, which reduces the influence of ambient lighting while keeping the necessary image details since the contour of the scene is the stable and unchanged feature. Also, we use a Fourier analysis to explain the contribution of different frequency spectrum in the correlation. By using this frequency information, we propose a simpler kernel filter method based on pre-process, which has the similar performance with non-linear filter pre-process but has less computation complexity. This simple kernel is more suitable for the embedded DSP real-time application.
Automatic recognition of ship types from infrared images using superstructure moment invariants
Automatic object recognition is an active area of interest for military and commercial applications. In this paper, a system addressing autonomous recognition of ship types in infrared images is proposed. Firstly, an approach of segmentation based on detection of salient features of the target with subsequent shadow removing is proposed, as is the base of the subsequent object recognition. Considering the differences between the shapes of various ships mainly lie in their superstructures, we then use superstructure moment functions invariant to translation, rotation and scale differences in input patterns and develop a robust algorithm of obtaining ship superstructure. Subsequently a back-propagation neural network is used as a classifier in the recognition stage and projection images of simulated three-dimensional ship models are used as the training sets. Our recognition model was implemented and experimentally validated using both simulated three-dimensional ship model images and real images derived from video of an AN/AAS-44V Forward Looking Infrared(FLIR) sensor.
SAR image segmentation with level set and clonal selection algorithm
Xiaojin Hou, Shuang Wang, Licheng Jiao, et al.
Synthetic aperture radar (SAR) image segmentation is a fundamental problem in SAR image interpretation. SAR images often contain non-texture object and texture object. Level set method, known as deformable model, is a powerful image segmentation technique. It can get accurate contours of non-texture object, but has poor performance in getting contours of texture object. In this paper, a new modified model of level set based on clonal selection algorithm is proposed. We use clonal selection algorithm to choose some pixels near the contour, and then perform a neighborhood modification on the level set function during its evolution. The region texture information, supervising the modification process, is incorporated into the level set framework. This new method is particularly well adapted to detection of texture object of interesting. We illustrated the performance of the new method on SAR images. Furthermore, we compared our method with level set method and the modified model of level set based on standard genetic algorithm (SGA) in texture object detection results and image segmentation results. The experimental results show that incorporating region texture information into the level set framework, consistent texture objects are obtained, and accurate and robust segmentations can be achieved.
Image denoising using improved adaptive proportion-shrinking algorithm based on second generation bandelets
Biao Hou, Haigang Li, Licheng Jiao, et al.
As one important multiresoltion geometry analysis tool, second generation bandelets can make full use of intrinsic geometry regularity of images, and then produces a sparse representation. This paper proposes a new denoising method, which is based on second generation bandelets and improved adaptive proportion-shrinking algorithm. Experiments on natural images with additive Gaussian white noise show that our method not only has the high peak signal to noise ratio(PSNR) value, but also has finer impression in vision, especially, has better performance on preservation of edges information and textures information than the classical proportion-shriking algorithm.
New method for text detection and segmentation from complex images
Fang Liu, Xiang Peng, Tianjiang Wang
Textual information contained in images is a valuable source of high-level semantics for image indexing and retrieval. This paper proposes a new method to detect and segment text from complex images. First, a density-based clustering method is employed to discover the candidate text regions. The clustering method is from data mining area. It computes the density distribution of overall image and makes spatial connective pixels with similar color/grayscale into one region. The clustered regions are deemed as candidate text regions. Then simple heuristics are applied to delete those obvious non-text regions from the candidate. But there still exits a few non-text regions in the candidate. Therefore a texture-based method is used to select text regions from the filtered candidate text regions. Considering the time complexity of density computation in clustering step, an approximate algorithm is designed to improve the efficiency. Experimental result shows the method is robust to variations in text font, orientation, language, and size.
Computational simulation of second order motion perception
Bin Sun, Nong Sang
Physiological studies have provided clear evidence of neurons sensitive to second-order motion, and first-order motion mechanisms are blind to second-order motion. In this paper, we propose a computational simulation of second order motion perception, which bases on energy-based detector with a preceded nonlinear process called texture grabber. Generally, a texture grabber consists of a linear spatial filter, a linear temporal filter and a nonlinear transform, such as full-wave rectification. Here Difference of Gaussians (DoG) functions are used as the spatial filters, and Difference of Gammas (DoGamma) functions are chosen as the temporal filters. A series of experiments are computed and the results confirm that our motion perception system detects second-order motion correctly.
Target detection based on hierarchical saliency feature model
Yuehuan Wang, Qian Lan, Tianxu Zhang
A target description model based on hierarchical saliency feature and target detection based on the model are proposed, which integrates bottom-up and top-down vision attention mechanisms together. Some saliency features of target are extracted on multi-scale, such as local symmetry, corner and so on, which are then processed hierarchically. This makes detection process much simpler and robust. Experiments demonstrate that the approach proposed is effective.
Intrackability theory and application
Zheng Li, Haifeng Gong, Nong Sang, et al.
Many vision tasks can be posed as Bayesian inference, and the entropy of the posterior probability is a measure for uncertainty of perception, imperceptibility. In this paper, we studied the imperceptibility of multiple object tracking, intrackability. Entropy theory and Bayesian framework are used to represent multiple objects intrackability. Intrackability is computed by different kinds of tracking features. Feature selection is crucial for intrackability computation. An example of umbrellas tracking is shown in this paper. The intrackability which is computed by appearance and shape feature is compared. At last, we use intrackability to guide one application--Automatic grouping. Objects are dynamically merged and tracked as a group when they come close to each other. Automatic grouping reduces the representation when some details can't be perceived. After the intrackable part of the representation is discarded, the computation is reduced.
Scene understanding with tri-training
Lin Zhu, Jie Zhou, Jingyan Song
Scene understanding needs not only detecting objects in the scene, but also obtaining the relationship between the objects and the scene, for example the reasonable size and occurrence possibility of objects at one position in the scene. With this relationship, the traditional object detection approach, which may misclassify objects with wrong sizes or position of the scene, can be greatly improved. In this paper, a novel scale model is proposed to describe the understanding of the scene. The scale model consists of the occurrence possibility and the reasonable size of pedestrian in each position of the scene. The scale model is learned by counting the pedestrian examples with different sizes in different positions of the scene for a period of time, instead of computing the geometry and viewpoint information in a single image. The examples are detected automatically by a detector which is trained with tri-training based semi-supervised approach. Experimental results indicate that the scale model of the scene can be learned with semi-supervised detection without the information of the 3D geometry and the assumption of plain ground.
Visual tracking system for water surface moving targets
Lin Zheng, Michael Beetz, Suat Gedikli
Water surface moving targets tracking is a challenging problem in the field of computer vision. Because moving targets are in a cluttered environment and are occluded randomly by splashed water, it is difficult to accurately extract and track them. In this paper, by analyzing water surface's color and motion statistical characteristics, a two-step segmentation algorithm is proposed to extract these targets. Then a multi-view tracking systme is established to estimate the 3D trajectory of moving targets' center. We employ this system to canoe competition, and to compare our result with the standard 3D trajectories, which can be calculated by using the markers on the canoes. THe experiments show that the root median square error between our trajectories and the standard ones is very low.
Use of binary logistic regression technique with MODIS data to estimate wild fire risk
Hong Fan, Liping Di, Wenli Yang, et al.
Many forest fires occur across the globe each year, which destroy life and property, and strongly impact ecosystems. In recent years, wildland fires and altered fire disturbance regimes have become a significant management and science problem affecting ecosystems and wildland/urban interface cross the United States and global. In this paper, we discuss the estimation of 504 probability models for forecasting fire risk for 14 fuel types, 12 months, one day/week/month in advance, which use 19 years of historical fire data in addition to meteorological and vegetation variables. MODIS land products are utilized as a major data source, and a logistical binary regression was adopted to solve fire forecast probability. In order to better modeling the change of fire risk along with the transition of seasons, some spatial and temporal stratification strategies were applied. In order to explore the possibilities of real time prediction, the Matlab distributing computing toolbox was used to accelerate the prediction. Finally, this study give an evaluation and validation of predict based on the ground truth collected. Validating results indicate these fire risk models have achieved nearly 70% accuracy of prediction and as well MODIS data are potential data source to implement near real-time fire risk prediction.
Analysis of the tropical forest vegetation change in Xishuangbanna of P.R. China by using LANDSAT TM
Cunjian Yang, Rong He, Shanxi Luo, et al.
The tropical forest change of Xishaungbanna in P.R.of China was studied here, which includes several steps. Firstly, The LANDSAT TM images acquired respectively on March 26, 1987, February 10, 1997, and March 6, 2000 were geometrically corrected, matched, and radiantly corrected. Secondly, the Normal Different Vegetation Index (NDVI) was formulated for the three periods. Thirdly, the false color composition images were achieved respectively for the three periods, which were used for visually interpreting forest and non-forest. Fourthly, the forest areas of the three periods were extracted semi-automatically from NDVI of the three periods by their suitable threshold. Fifthly, the forest areas of the three periods were changed into vector form. Sixthly, the extracted forest area of the three periods in vector form superposes their false color composition images for evaluating their accuracies. Finally, the map of forest change among the three periods was made, and the forest change area was calculated. The research shows: The decreased tropical forest area between 1987 and 1997 is 234558 hectares, accounted for 19 percent of the tropical forest area in 1987. The decreased tropical forest area between 1997 and 2000 is 41237 hectares, accounted for 4.2 percent of the tropical forest area in 1997
Modeling human false alarms using clutter metrics
Honghua Chang, Jianqi Zhang, Delian Liu
The TSSIM clutter metrics correlate amazingly well to both the experimental detection probabilities and the mean detection time. Based on the analysis of both probabilities of the correct detections and of the total (correct and false) number of detections made by human observers, a mathematical formula for predicting the probability of false alarms as a function of clutter metrics is presented in this paper. Comparing real experimental data with the predicted products reveal very good agreement, which is very helpful in understanding human behavior mechanisms regarding target detection tasks. It is concluded that the human observer behaves as fixed threshold signal processor /Non-CFAR.
Real time object tracking using adaptive Kalman particle filter
Lin Gao, Peng Tang, Zhifang Liu
In this paper, a visual object tracking algorithm based on the Kalman particle filter (KPF) is presented. The KPF uses the Kalman filter to generate sophisticated proposal distributions which greatly improving the tracking performance. However, this improvement is at the cost of much extra computation. To accelerate the algorithm, we mend the conventional KPF by adaptively adjusting the number of particles during the resampling step. Moreover, in order to improve the robustness of tracker without increasing the computational load, another two modifications is made: firstly, the covariance matrix of Gaussian noise in the dynamic model is dynamically updated according to the accuracy degree of the prediction. Secondly, the similarity measurement is performed by a scheme that adaptively switches the likelihood models. Experimental results demonstrate the efficiency and accuracy of the proposed algorithm.
Stochastic approach based salient moving object detection using kernel density estimation
Peng Tang, Zhifang Liu, Lin Gao, et al.
Background modeling techniques are important for object detection and tracking in video surveillances. Traditional background subtraction approaches are suffered from problems, such as persistent dynamic backgrounds, quick illumination changes, occlusions, noise etc. In this paper, we address the problem of detection and localization of moving objects in a video stream without apperception of background statistics. Three major contributions are presented. First, introducing the sequential Monte Carlo sampling techniques greatly reduce the computation complexity while compromise the expected accuracy. Second, the robust salient motion is considered when resampling the feature points by removing those who do not move in a relative constant velocity and emphasis those in consistent motion. Finally, the proposed joint feature model enforced spatial consistency. Promising results demonstrate the potentials of the proposed algorithm.
Aircraft type recognition based on convex hull features and SVM
Yuan Liu, Xiuqin Wu, Richang Hong
Most current algorithms of aircraft type recognition are based on the binary images which are obtained by utilizing the technology of image segmentation. Thus the effect of image segmentation will influence the sequent classification to a great extent. Moreover, image segmentation in complex background remains a challenging research area. In our work, we propose a novel aircraft type recognition algorithm based on the aircrafts' convex hull features and Support Vector Machine (SVM). We first obtain the aircrafts' external contours while removing background. And then, we compute the planar convex hulls of the external contours. Based on the convex hulls, we combine the characteristics unique to the aircraft object, to introduce an extracting method of major symmetry axle and corresponding characters. Finally, we select the SVM which has high generalization capabilities and high performance in tackling small sample size in the pattern classification task to perform the classification. Experiment results show that the convex hull feature of aircraft object is approximately invariant, and can successfully eliminate the need to segment the object region from the complex background. The aircraft type recognition is efficient and feasible, and especially applicable for raw gray images.
Reduced memory zerotree envelop coding for wavelet image
Wentao Wang, Guoyou Wang, Jianguo Liu
In this paper, we propose a new image codec, which called embedded zerotree wavelet coefficients envelop coding (EZWCEC). The coefficients envelop is characterized by describing the global tendency of the significant wavelet coefficients. Based on the empirical analysis and experimental results, our EZWCEC algorithm restores the trend of the significant wavelet coefficients and estimates the magnitude of some insignificant coefficients on the decoder. Unlike other zerotree coding algorithms such as Said and Pearlman's SPIHT using three lists, EZWCEC only uses two lists during encoding and decoding. So the memory requirement for the hardware implementation is reduced significantly. Although PSNR values for EZWCEC are lower than SPIHT's and JPEG2000's, our experiment results have shown that EZWCEC can dramatically improve the visual quality of reconstructed at low bit rates (e.g., below 0.1bpp).
Improved clustering algorithm for image segmentation based on CSA
Xiaohua Zhang, Pu Yang, Licheng Jiao, et al.
Image segmentation is the prerequisite step for further image analysis. Segmentation algorithms based on clustering attract more and more attentions. In this paper, an image-domain based clustering method for segmentation, called CSA-CA, is proposed. In this method, a scale parameter is introduced instead of an apriori known number of clusters. Considering that adjacent pixels are generally not independent of each other, the spatial local context is took account into our method. A spatial information term is added so that the near pixels have higher probability to merge into one cluster. Additionally, a clonal selection clustering operator is used so that a cluster is capable of exploring the others that are not neighboring in spatial but similar in feature. In the experiments we show the effectiveness of the proposed method and compare it to other segmentation algorithms.
Segmentation of serial CT images based on an improved Mumford-Shah model
Yufei Chen, Weidong Zhao, Zhicheng Wang
Segmentation of medical image is an indispensable process in image analysis and recognition, and it provides the basis of quantitative analysis of images about human organs and functions. The Mumford-Shah model using level set method is more robust than other curve evolution models to detect discontinuities under noisy environment, which has been widely used in the field of medical image segmentation. Consequently, serial computed tomography (CT) image segmentation algorithm based on an improved Mumford-Shah model is presented. First of all, the window transformation technique of medical images is introduced, which is able to display the digital imaging and communications in medicine (DICOM) images directly and distinctly with a little information loss. Secondly, the characteristics of serial CT images as well as the topological structure relation between them are analyzed, followed by the processing method of CT image sequence, which can make the serial CT image segmentation much more automatically and swiftly. Thirdly, in the light of the problems of segmentation speed and termination in traditional Mumford-Shah model, a novel segmentation algorithm based on image entropy and simulated annealing is presented. The algorithm alleviates these two problems by using the image entropy to displace the energy coefficients in the original energy function, and also combining the simulated annealing to terminate the contours evolution automatically. Finally, the algorithm is applied in some experiments to deal with serial CT images, and the results of the experiments show that the proposed algorithm can provide a fast and reliable segmentation.
Precise tracking algorithm for small target based on event supervision
Yun Li, Tianxu Zhang
A precise tracking algorithm for small target based on event supervision is introduced in this paper. The target chains and object aggregation are established firstly, Tri-level scan filter contains grey intension filter, shape filter and location filter, is adopted to implement data relevancy between target chains and object aggregation from coarse to fine. Movement trend of object to tracking target which is classified into follow, approach and leaving, and events include envelop and combination are detected, supervised and processed. On the other hand, based on the analysis of the error model for target centroid estimation, a recursive approach method for target centroid calculation with high precision is adopted in the algorithm. It combines tracking, recognition and prediction effectively base on the tracking theory of human eyes. Experimental results show that the method is feasible and effective.
High-accurate line feature extraction algorithm based on line diffusion function model
Shunyi Zheng, Guozhong Su, Jianqing Zhang
A high-accurate line feature extraction algorithm based on line diffusion function model has been proposed in this paper, using local gray value variations to precisely identify edge locations. This paper firstly analyzed the primary principle of line diffusion function model, then showed that how to use the information of image window to determine the exact position of line features in images. During the process, the mathematical model is provided and high-accurate line feature extraction algorithm has been developed. Finally, this algorithm has been used to extract line features from aerial images of urban area and experimental results are presented to demonstrate the efficiency and accuracy of the proposed algorithm.
Multi-object segmentation algorithm based on improved Chan-Vese model
Xiaowei Fu, Mingyue Ding
Because Chan and Vese(C-V) model using one level set function can only represent one object and one background, it cannot represent multiple junctions of multiple objects. In this paper, an improved multi-object segment algorithm is proposed based on C-V model of single level set. First, the given image resolution is deduced by wavelet transform. Since the low resolution approximate image contains less noise and pixels, it can speed up the active contour evolution. Secondly, an improved C-V model of a single level set is introduced to obtain the multi-objects' approximate contour, which can make use of topology split information of the contour effectively. Thirdly, the inverse discrete wavelet transform is used to the resulted image and level set of the coarse scale image, which can get the approximation contour on the original image. Lastly, the approximation contour is taken as an initial level set function and the second active contour evolution is performed on the original image to get the real multi-objects contour. Experimental results show that the proposed algorithm can realize the multi-object segmentation effectively and quickly.
Multiple states and joint objects particle filter for eye tracking
Jin Xiong, Huanqing Feng
Recent works have proven that the particle filter is a powerful tracking technique for non-linear and non-Gaussian estimation problem. This paper presents an extension algorithm based on the color-based particle filter framework, which is applicable for complex eye tracking because of two main innovations. Firstly, an employment of an extra discrete-value variable and its associated transition probability matrix (TPM) makes it feasible in tracking multiple states of the eye during blinking. Secondly, the joint-object thought used in state vector eliminates the distraction from eyes to each other. The experimental results illustrate that the proposed algorithm is efficient for eye tracking.
Algorithm of optical remote sensing image registration based on strong edge region
Ting Yao, Dong Yin
Aiming at the registration of optical remote sensing images, an algorithm based on strong edge region is proposed. First, the strong edge regions are extracted. Then combines with the regions' moment invariants and RANSAC method, it can obtain an accurate match of the strong edge regions. Utilizing the centroids of the matching regions as control points in the affine geometric distortion, an automatic registration is performed. A large number of experiments are fulfilled with SPOT and Quickbird satellite images and good results are obtained.
Segmentation-based reflectance recovery
Xiangyang Wu, Hongxin Zhang
The reflectance properties of a surface are an essential factor in its appearance. Much previous work has focused on the problem of reflectance recovery from images. These methods must assume an a priori grouping of pixels into uniform-reflectance regions. In this paper we presented a method for automatic grouping of pixels for reflectance estimation. First a over-segmentation is achieved by traditional image segmentation .For each image region of the over-segmentation, a probability distribution is built and a reflectance subspace is formed by likelihood thresholding. The regions with the same reflectance are then merged by adapting a traditional bayesian formulation for image segmentation to increase estimation accuacy. After completing the merging process, reflectance parameter estimates are computed for the remaining subspaces by the maximum likelihood reflectance estimate.The experiment results on a synthetic scene and a real scene show our method can achieve a more accurate image segmentation and reflectance estimation than traditional methods.
Fingerprint image segmentation based on multi-features histogram analysis
Peng Wang, Youguang Zhang
An effective fingerprint image segmentation based on multi-features histogram analysis is presented. We extract a new feature, together with three other features to segment fingerprints. Two of these four features, each of which is related to one of the other two, are reciprocals with each other, so features are divided into two groups. These two features' histograms are calculated respectively to determine which feature group is introduced to segment the aim-fingerprint. The features could also divide fingerprints into two classes with high and low quality. Experimental results show that our algorithm could classify foreground and background effectively with lower computational cost, and it can also reduce pseudo-minutiae detected and improve the performance of AFIS.
Improved method of nonlinear anistropic diffusion filtering based on open system
Weixin Wu, Hongchen Liu
We propose an improved P-M nonlinear filter based anisotropic diffusion. In the new method, an open diffusion system is adopted to improve edge preservation in image denoising process. In the diffusion system, we choose part of pixels as "origin" pixels and "convergence" pixels. Value of these pixels does not change during filtering process. As a result, system energy exchange exists both within the system and between system and outer space. After filtering process, part of the noise energy is filtered out of the system. We test the performance of the new method and compare the result with Gaussian, P-M and Catte filter. Experimental results demonstrate the effectiveness in edge preservation of the new method.
Automatic extraction of water body based on EOS/MODIS remotely sensed imagery
Qiuwen Zhang, Cheng Wang, Fumio Shinohara
As a new type of remotely sensed resources, EOS/MODIS imagery has moderate spatial resolution, high temporal resolution and high spectral resolution. Based on the spectral character analyses of water body in EOS/MODIS images, this paper proposes models of water body extraction from EOS/MODIS remotely sensed imagery. It shows that the spectral differences between water body and other targets in a single channel of EOS/MODIS image are not very clear; however, their spectral relationship differences between different channels of EOS/MODIS images are obvious. Water body can be segmented and extracted by the criteria of NDWI<-0.1 or NDVI<0.04 & (CH4-CH5)>2.0. The precisions of water body extraction with the established models are up to 90%.
Detection algorithm of unattended package in surveillance sequence based on spatial-temporal accumulative frame difference pictures
Xianyi Ren, Hongjun Chen, Tianhuai Ding, et al.
Detecting unattended packages in scenes has several applications in video surveillance. This paper proposed an algorithm for automatically detecting unattended packages in surveillance video. This paper proposed an algorithm for automatically detecting unattended packages in surveillance video using information available in the special-temporal accumulative difference picture. The proposed algorithm is able to detect abandoned objects in the surveyed environment and then to alert a human operator whenever a dangerous situation is recognized. The algorithm can handle situations where the background of the scene is cluttered and not completely static but contains small motions such as tree branches and bushes. The experimental results show that the algorithm is able to identify static objects in real video sequences in real time and it is very promising.
Study on edge extraction methods of the droplet image profile
Sijia Zhu, Han Yan, Wenqian Wu
It is a key technology to detect the edge of the droplet profile exactly for application in drop volume calculation, liquid target identification and liquid characteristics analysis. The droplet images in various stages during the drop growth are firstly acquired and stored real-timely through ICCD. After a series of preprocessing, including image cutting, noise filtering, image segmentation and filling, a proper edge detection method is chosen to extract the droplet image profile. The principle and algorithm of some classical and newly developed methods for edge detection are introduced and compared in detail, such as differential operator, Laplacian operator, improved Canny operator, wavelet transformation, mathematical morphology, fuzzy operator and fractal geometry. Wavelet transformation can be used effectively for extracting the droplet image profile, because of its advantages of multiple resolution, focus variation and edge enhancement. The image records of the droplet formation and the detected profile curves are presented.
Automatic recognition of airport in remote sensing images based on improved methods
Yan Chen, Kaimin Sun, Jingxiong Zhang, et al.
Airport recognition remains an important and challenging topic for research. In general, there are four main steps in the process of airport recognition: pre-processing of remote sensing images, features extraction, rough location and the recognition of airport. This paper puts forward an automatic airport recognition method which adopts improved methods in each step. In pre-processing, an edge-preserve image smoothing algorithm based on Convexity Model is developed. In features extraction, the Canny operator and chain codes are used. An improved Κ-Means lines segmentation and estimate rules are used to find candidate areas in rough location of the airport. And those candidate regions are binarized and prior knowledge is used in airport recognition. Experimental data and application results show that the above methods are efficient and enhance the accuracy in airport recognition.
Novel motion object extraction algorithm
Yi-Jun Xiao, Bang-Pin Wang, Xiao-Chun Lu, et al.
In intelligent transport system, it is very important to precisely segment motion object from complex scene. Background difference and frame difference are two classic motion object extraction algorithms. If there are shadows associated to moving objects, both of the methods cannot extract moving object precisely. With this problem, this paper proposes a motion object extraction algorithm based on active contour model (ACMMOE). The following steps are performed in ACMMOE: firstly, moving areas involving shadows are segmented with classical background difference algorithm. Secondly, perform shadow detection and coarsely removal, then using grid method to extract initial contours. Finally, use active contour model approach to the contour of the real object by iteratively tuning the parameter of the model. The proposed algorithm adopts greedy algorithm to speed up the solution to active contour model. Experiments show the algorithm can remove the shadow and keep the integrity of moving object.
Color image segmentation by integrating texture measure into JSEG method
Qinghong Sheng, Hui Xiao
We present a new color image segmentation method that combined texture measures and the JSEG (J measure based JSEGmentation) algorithm. In particular, two major contributions are set forth in this paper. (1) The two measures defined in JSEG and the Laws texture energy is discussed respectively and then we find that the over-segmentation problem of JSEG could be attributed partly to the absence of color discontinuity in the J measure. (2) A new measure is proposed by integrating the Laws texture energy measures into the J measure, on which our segmentation method is based. The new segmentation method taking account of both textural homogeneity and color discontinuity in local regions can be used to detect proper edges at the boundaries of shadows and highlights. Performance improvement due to the proposed modification was demonstrated on a variety of real color images.
Novel adaptive multi threshold image segmentation algorithm
Hong Jiang, Zhang Ren
A novel adaptive multi threshold image segmentation algorithm is proposed in this paper. This proposed segmentation algorithm has two unique characteristics: it fits the 1-D graylevel histogram of the image by potential base function and thereby adaptively determines the classification number by potential function clustering; based on the graylevel co-occurrence matrix, it acquires the multi segmentation thresholds which makes the shape connectivity maximum according to the shape connectivity criterion. Both theoretical analysis and simulation results indicate that the performance of this new adaptive multi threshold segmentation algorithm is superior to those of the conventional threshold segmentation algorithms. And it has not only a low computing cost, but also shows quite good segmentation effect. Besides, it is insensitive to noises and interferences.
Method to reduce over-segmentation of images using immune clonal algorithm
Jianhua Liang, Shuang Wang, Licheng Jiao
Image segmentation is a difficult task. These years, researchers have proposed many segmentation methods based on Evolutionary Algorithms, but most of them used Evolutionary Algorithms to optimize the parameters of an existing segmentation algorithm. This paper tries to use the Evolutionary Algorithms to segment images expecting to explore a new way of image segmentation. The method described in the paper pre-segments the image by Watersheds and then merges it by Immune Clonal Algorithm (ICA). To implement the task, several operators are proposed such as the DC operator, the Proportional Creation of the First generation operator, and fitness function based on JND and average gray value. In the end, the proposed method is compared with another method using GA. The experiments show that the method is effective and the work is significant.
Novel color image segmentation using self-generating prototypes
Chunping Liu, Xiaohua Yuan, Zhaohui Wang
A new self-generating prototypes method based on SGNT is presented. This method uses reference patterns as initial prototype. This procedure can be implemented in a SGNT with specific architecture consisting of one root and the initial class number of reference patterns. The leaf in SGNT is defined with prototype vector, learning vector, center property vector and distant property vector. After training, prototype set are outputted. The main advantage of this method is that both the number of prototypes and their locations are learned from the training set without much human intervention. Experiments with synthesis and real color image the excellent performance of this classification scheme as compared to existing K-nearest neighbor (K-NN) and Learning vector quantization (LVQ) algorithm.
Edge detection of license plate based on wavelet transform and quantum genetic algorithm
Yourui Huang, Hong Duan
Edge detection is one of the most commonly used operations in image analysis of license plate. The classical edge detection algorithm based on wavelet transform utilizes a threshold to remove noise from license plate image, and then detects edge with wavelet transform. In the condition with strong noise, the classical edge detection algorithm often works not well enough. The proposed edge detection algorithm combines the wavelet transform and quantum genetic algorithm and significantly improves the result. Specifically, wavelet decomposition is applied to get a set of low and high frequency sub-images from image of license plate and quantum genetic algorithm is applied to optimize wavelet denoising threshold which removing the noise of the high-frequency sub-images, finally the edge image is obtained by reconstructing the high-frequency sub-images without noise. The efficiency of the method is better than the classical one and is proved by computer simulation.
Steganalysis method based on SVM and statistic model in contourlet domain
Hongping Xu, Jianhui Xuan
There is an urgent demand on steganalysis, which analyzes if an image includes hidden information, or further decodes the hidden information. This paper proposes a steganalysis method based on statistics model in contourlet transform domain. The proposed method is a blind universal steganalysis method, which does not aim at specified steganography method. The contourlet coefficients of natural image shows obvious regularity which includes sparsity and clustering in subband and similarity across scales. The popular statistical model in wavelet subband is the generalized Gaussian distribution (GGD) model, which can capture the first-order statistical features in subband. While the GGD model can not characterize the dependency between coefficients. The proposed steganalysis method takes contourlet statistics in subband and dependency between contourlet coefficients into account. The dependencies are measured using mutual information. The selected features include parameters of GGD model in subband, the mutual information between coefficients. The classificator chose is Support Vector Machine (SVM). The experimental results show that the features used in the proposed method are valid, when the dependencies between contourlet coefficients are taken into account, the false positive rate is greatly lower than the case in which the dependencies are not considered.
Preliminary study on ice crevasse texture analysis and recognition
Chunxia Zhou, Dongchen E, Zemin Wang
The Antarctic is in very close relationship with the global climate, ecology environment, and the future of the human being. And it is unscientific to explore the Antarctic without any touch. While, crevasse is one of the most dangerous factors to the team members during the field expedition. Crevasse detection is very important in polar scientific research expedition for the safety; meanwhile, it is also meaningful information for ice flow monitoring. This paper presents the preliminary study on ice crevasse texture analysis and recognition based on SPOT image and coherence map derived from SAR image of Grove Mountains, east Antarctica. Since radar can penetrate the snow, it can detect the crevasse under the snow which can't be detected by optical satellite data. Based on the texture characteristics, gray level co-occurrence matrix is chosen at first to recognize the crevasse in SPOT image and coherence map respectively. And the results and the difference are analyzed. Optical and radar imagery both are valuable, however, there is no single sensor that gives 100 percent of the crevasses. Meanwhile, gray level co-occurrence matrix method can not detect the crevasse at 100 percent accuracy. More texture analysis method will be studied in further research.
Descreening method of scanned halftone image based on wavelet
Yaohua Yi, Juhua Liu, Changhui Yu, et al.
In the workflow of image reproduction, the halftone image is usually used as scanned original image. The moiré patterns are a result of intervention during the process of scanning. This paper provides a descreening method of scanned halftone image based on wavelet according to the cause and the character of moiré patterns. Experiment result shows that this descreening method can remove the moiré patterns effectively and at the same time hold the definition and entropy of the scanned halftone image.
Moving point target trajectory detection based on dual-Hough transform
Jing Hu, Tian-Xu Zhang
This paper deals with research on detecting moving point target trajectory in image sequence. A novel method is presents for this purpose, which combines two 2-dimension Hough transforms to suppress noise points and to detect trajectory points in time order. The first Hough transform has an accumulators array using a restricted voting process and a set of straight lines are found in the image plane. A new T-L parameter space is proposed which is derived from these straight lines. In the second transform, collinear points are mapped into T-L space and it is easy to find the direction of motion. Experimental results show that our method can effectively extract moving point target trajectory accurately in a limited observing time especially scanning images from large numbers of noise points while search region is much larger than target movability.
Image quality assessment method in intelligent transportation systems
Bangping Wang, Jian You
In modern ITS (Intelligent Transportation Systems), the close shot images captured by camera are used to precise recognition of information of vehicles such as VLP (vehicle license plate), VS(vehicle shape) , VBC(vehicle body color) and etc. The precise recognition of vehicle information seriously depends upon quality of images captured by camera. The assessment of image quality is a meaningful work, which can be used to monitor the working state and adjust the control parameters of camera, further more can guide the recognition of information of vehicle. This paper proposes a novel content-based method of assessing images quality for close shot ones in ITS . The method is objective image quality assessment without reference image, which is point to single image. The assessment includes distortion type and distortion amount. Experiments show the method is valid and robust.
Speckle noise reduction based on the theory of rough set and entropy
Jun Li, Guohua Chen, Tiejun Ma
A new effective algorithm of speckle noise reduction was presented in the paper, which took the characteristics of the multiplicative of speckle noise and the complexity of fringe pattern into account explicitly. The relationship expressions are deduced between intensity and speckle noise from the intensity distribution of speckle pattern, which indicates that the intensity ratio is equivalent to the speckle noise ratio in a neighborhood. Due to the complexity and correlativity of fringe pattern information, it leads to the problems of uncertainties during the information processed, the algorithm of speckle noise reduction was built based on exponent entropy, and the approximation precision based on rough set was considered in order to preserve the detail information very well. The presented method has a preferable performance based on the integrated visual and quantitative comparison. Because the filter parameters are adaptively determined by the speckle noise coefficient in the local window, the presented method was able to remove the noise in fringe pattern, and simultaneously preserve its edge effectively. The presented algorithm is effective for the laser speckle noise of fringe pattern particularly.
Face tracking based on grey prediction
Zhiyu Zhou, Jianxin Zhang
Although single skin color model is the normal method of face detection, it often has the disadvantages such as "ultra-detection" and bad real-time performance. This paper presents a new method of face tracking based on grey prediction model GM(1,1) combined with skin color model. The method uses grey prediction to minish the region of matching search for skin color model and the matching result to update the prediction basis of the grey model. In order to acquire the initial position of face, moving information of face is used, and the moving region is automatically acquired with information entropy. Compared to the α-β-γ filtering with assumption that object in image sequence makes uniformly accelerated motion, the experiment results of this method show that grey prediction model GM(1,1) can maintain minor error stably. The result of grey prediction is closer to the real motion trajectory, and better reflects the motion trend of face. It greatly enhances such two important indexes as robustness and real-time performance under the system tracking process.
Spatial variability analysis to remote sensing image on different scales with wavelet variance
Lingling Ma, Lingli Tang, Zhaoliang Li
Nowadays, Remote sensing is widely applied into flood monitoring. In order to obtain the accurate flooded area, water body need be well segmented, however, image segmentation is still a hard problem. In fact, each image segmentation method has its advantages and disadvantages, single method is hard to acquire satisfactory results, so two or more methods combination are applied. In this paper, the algorithm of combination watershed transformation and region merging based on morphologic gradient is introduced and applied in a typical flooded region, Quyuan town of Hunan province of China to extract the area of target water bodies, at the same time, other segmentation methods: ISODATA and the combination segmentation methods of optimal threshold and mathematical morphology, are also applied. Finally, compared with actually measurement area of the target water body by Differential GPS, some suggestive conclusions are drawn.
Effective segmentation method for water body: the algorithm of combination watershed transformation and region merging based on morphologic gradient images
Yugang Tian, Gang Chen
Nowadays, Remote sensing is widely applied into flood monitoring. In order to obtain the accurate flooded area, water body need be well segmented, however, image segmentation is still a hard problem. In fact, each image segmentation method has its advantages and disadvantages, single method is hard to acquire satisfactory results, so two or more methods combination are applied. In this paper, the algorithm of combination watershed transformation and region merging based on morphologic gradient is introduced and applied in a typical flooded region, Quyuan town of Hunan province of China to extract the area of target water bodies, at the same time, other segmentation methods: ISODATA and the combination segmentation methods of optimal threshold and mathematical morphology, are also applied. Finally, compared with actually measurement area of the target water body by Differential GPS, some suggestive conclusions are drawn.
Accurate dynamic scene model with double-layer Gaussian mixture
Hong Yang, Yihua Tan, Jinwen Tian, et al.
Gaussian mixture model is a popular method to model dynamic scenes viewed by a fixed camera. However, we found that it is not a trivial problem for each component of Gaussian mixture to learn the accurate parameters for complex pixels. Furthermore, traditional method of Gaussian mixture has to make a tradeoff between system stability and convergence rate. We developed a mechanism of double-layer Gaussian mixture model for moving object detection from dynamic scenes, which can improve the convergence rate without compromising the system stability. Additionally, temporal consistency of variances was taken into account to alleviate camouflage problems in the process of detection.
Adaptive Wiener filter based on wavelet transform
Wenruo Bai, Baofeng Zhang, Qianqian Bai
This paper presents a new method for image filtering process, an adaptive Wiener filter, which is based on wavelet transform. The filter provided uses threshold mode filtering as its precursor. The pertinence between wavelet coefficients is taken into consideration in the process of determining the threshold quantities, which makes the filter adaptive. The result of the threshold mode filter is then processed by Wiener filtering model. The whole image filtering system is composed of threshold mode filtering based on wavelet transform as well as Wiener filtering technique. What's more, the performance is also an important subject. And it gives a better smooth filtering result.
Overlaid caption extraction in news video based on SVM
Manman Liu, Yuting Su
Overlaid caption in news video often carries condensed semantic information which is key cues for content-based video indexing and retrieval. However, it is still a challenging work to extract caption from video because of its complex background and low resolution. In this paper, we propose an effective overlaid caption extraction approach for news video. We first scan the video key frames using a small window, and then classify the blocks into the text and non-text ones via support vector machine (SVM), with statistical features extracted from the gray level co-occurrence matrices, the LH and HL sub-bands wavelet coefficients and the orientated edge intensity ratios. Finally morphological filtering and projection profile analysis are employed to localize and refine the candidate caption regions. Experiments show its high performance on four 30-minute news video programs.
Reconstructed key frame and object motion based video retrieval
Shuangyan Hu, Junshan Li, Kun Li, et al.
This paper proposes a video retrieval scheme which can retrieve desired video clips from video databases using color and object motion. The retrieval method includes two steps. In the first step, get the Intra picture frames (I-frames) set from the query MPEG video and reconstruct the key frame of the video based on the set. Then, the video retrieval equals to the retrieval of the reconstructed key frame(R-key frame) and can be easily performed according the methods of content based image retrieval. The second step, the local object motion information that is local motion vector field, is extracted from the video clips set which is the result of the first step, and the final similarity of videos is measured based on the constructed directional histogram. Experimental results show that the proposed two-step retrieval method performed excellently for video retrieval.
Automatic bridges recognition in medium resolution SAR images based on multi-scale feature analysis
Dong Yin, Feng Zhang, Xin Chen, et al.
This paper handles with the problem of bridge recognition in Synthetic Aperture Radar (SAR) images. Based on features analysis of bridges, rivers and land in different spatial resolution SAR images, a method of multi-scale analysis is proposed. Firstly, for preventing noise, filtering the original medium resolution image is performed. And the image will be down-sampled to ten times. Secondly, rivers will be extracted automatically by dynamic programming in low resolution level and the bridge candidates' position will be obtained by apices locating. Finally, the mapped image will be cut out to become regions of interest (ROI). And the bridge target regions will be detected by using constant false alarm rate (CFAR). The example results indicate that the processing speed for bridge recognition can be greatly improved and the precision of recognition can also be ensured.
Edge detection operators applications in flame image recognition and processing
Based on region-segmentation method and flame image edge detection operator, flame image pre-processing and edge detection can effectively improve flame image processing efficiency. Relatively effective flame image edge detection operator is proposed in this paper. It provides basis for further flame parameters extraction.
Changes of spectrum of light scattering on quasi-homogeneous random media
Yu Xin, Yanru Chen, Qi Zhao, et al.
We study the spectrum of the field generated by beam radiated from quasi-homogeneous (QH) electromagnetic source scattering on QH media. Using the unified theory of coherence and polarization of random electromagnetic field, formulas for the spectral density and of the three dimensional scattered field for incidence of plane wave and a beam were derived respectively. From the spectral density of the scattered field, the normalized spectrum of the scattered field was obtained. It was found that the spectrum of the scattered field has a shift similar case of propagation.
Method for extracting specific parallel straight lines
Lingli Zhao, Jianjun Zhu, Shuai Liu, et al.
Targets based on linear feature often need to be extracted and identified with high precision in the application of photogrammetry. Linear features such as hydrological object boundaries, roads and the boundary of other man made objects are very important for geospatial information extraction and analysis from remotely sensed imagery. This paper deals with subpixel accuracy extraction of linear features, especially specific parallel straight lines. There are many algorithms for localization, such as Gray moment operator, Hough transformation algorithm Forstner operator and so on, just only adopting a single algorithm, the precision of localization is low, and the calculation is complicated. So a new mixed method is proposed in the paper and the procedure can be divided into two steps. Firstly, the rough location of the parallel straight lines was extracted with Hough transformation algorithm. We could get the initial value of parallel straight line in this step. Secondly, straight linear fitting based on Gray moment operator for edge location was adopted to extract the straight lines with high precision. The experimental results indicate that the mixed method for subpixel localization locates the target very validly, and some practicable conclusions are received.
Method of image enhancing based on shape feature of target
Wei Wu, Lan Xiang, Zude Zhou
It is difficult to get ideal effect that research on target enhancement and segmentation are developed by simply using features of edge, texture and gray in complex background. Studies on target enhancement and segmentation have always been the research focus for a long time. Considering known regular shape target in complex background, an approach of target enhancement and segment based on shape feature is proposed. According to analyze the shape of regular target such as line and arc, first, Top-hat is used to enhance target. Then, threshold segmentation method, thinning and deburring are following. Third, a new image is acquired by the Hough transform which can find points in edge image similar with target in shape and property. The very points would be reserved. Moreover, the new image acquired is the initial condition of reconstruction, the original image is limit condition, and target enhancement is achieved by grayscale morphological reconstruction. Finally, binaryzation processing of image was used to segment target. Experimental results demonstrate that the proposed algorithm has an encouraging performance.
Level set method in standing tree image segmentation based on particle swarm optimization
Jiangming Kan, Hongjun Li, Wenbin Li
For the intelligent pruning machine, a machine vision system is pre-requisite. Standing tree image segmentation is a key step for the machine vision system. An efficient scheme for tree image segmentation was proposed according to the need of the machine vision system of the intelligent pruning machine. The scheme is a level set method based on particle swarm optimization. According to principal of the level set method, the image segmentation is formulated as one of optimization problems. The energy function is taken as the segmentation quality criteria, which consists of an internal energy term that penalizes the deviation of the level set function from a signed distance function, and an external energy term that drives the motion of the zero level set toward the desired image feature, such as object boundaries. In this paper, the method used particle swarm optimization to solve the optimization problems that is different from the ordinary level set method that uses the partial differential equation method in some literatures. In experiments, tree images with different background are selected to test the efficiency of the scheme that presented in this paper. In order to test the antimonies performance of the scheme that presented in this paper, a tree image added Gaussian white noise is selected. From the results of the tree image segmentation, the scheme that presented in this paper is more efficiently. The experimental results demonstrate the scheme is more effective and time-saving than the ordinary level set method.
Auto-segmentation based on graphcut and template
Hong Zhang, Wei Zuo, Ying Mu
In this paper, the main idea is to use the prior knowledge to guide the segmentation. Firstly the continuity among adjacent frames is used to create a motion template according to the Displaced Frame Difference's (DFD) higher character . And then the color template is established by using the k-means clustering. Based upon the information derived from the previous two templates, the segmentation image is defined as foreground, background and boundary regions. Then, the segmentation problem is formulated as an energy minimization problem. The hard edge of foreground is then obtained by implementing graph-cut algorithm. Experimental results demonstrate the effectiveness of proposed algorithm.
Novel methods of image segmentation based on data field
Qizhi Xiao, Kun Qin, Zequn Guan
The paper introduces data field theory into the field of image segmentation, and proposes novel methods of image segmentation. Simulating some physical fields, such as electromagnetic fields, the fields of gravity, the fields of nuclear force etc., each data point is considered as a material particle with radiate ability, and the interaction of all data points forms a data field. Each image pixel is considered as a data point, all pixels in an image form an image data field which is the basis of image segmentation. The basic steps of image segmentation approaches based on data field include: (1) generate image data field; (2) optimize and adjust influence factor; (3) implement image segmentation based on the optimal image data field. Two kinds of methods are proposed in this paper. One is the method of interactive image segmentation: directly give the isopotential threshold value of the interesting target object, and implement image segmentation and extract the special object; the other one is the method of automatic segmentation: automatically segment the whole image based on the given classification number. Some experiments and analysis confirm the validity of the proposed methods.
Automatic event recognition and anomaly detection with attribute grammar by learning scene semantics
Lin Qi, Zhenyu Yao, Li Li
In this paper we present a novel framework for automatic event recognition and abnormal behavior detection with attribute grammar by learning scene semantics. This framework combines learning scene semantics by trajectory analysis and constructing attribute grammar-based event representation. The scene and event information is learned automatically. Abnormal behaviors that disobey scene semantics or event grammars rules are detected. By this method, an approach to understanding video scenes is achieved. Further more, with this prior knowledge, the accuracy of abnormal event detection is increased.
Ship detection in low-resolution SAR images based on background suppression
Weidong Yang, Tianxu Zhang, Yunsheng Liu
Owing to the capabilities of time and weather independent imaging, Synthetic Aperture Radar (SAR) is one of the most promising remote sensors for ship detection in the field of ocean surveillance. Low-resolution SAR imaging sensor provides wide area coverage, which is useful for monitoring large expanses of coastal waters for the presence of ships. In low-resolution SAR images, the ship detection belongs to small target detection problem in speckle noise image. The background suppression method against the typical features of generic non-homogeneous sea clutter and the strong background noise should be a crucial step in the object detection. A background inhibition network model is firstly built on the theory about the lateral inhibition of biological visions. Secondly, a novel region of interest (ROI) extraction method based on the Renyi's entropy is proposed. Finally, a homogeneity-testing method to reduce the false ship target of ROI is suggested, which was by the means of statistical distribution function of intensity variance-mean ratio (VMR) in homogeneity. The experimental results demonstrated that the lateral inhibition method can not only suppress the background and clutters well, but also can enhance the ship target perfectly. The proposed algorithm has good performance in adapt to various sea-condition and observing angle.
Improved thresholding method based on Tsallis-Havrda-Charvat entropy
Jingjing Chen, Jin Wu
This paper presents a thresholding method for image segmentation by using an improved thresholding output function on a two-dimensional (2-D) histogram based on Tsallis-Havrda-Charvat entropy principle. The Tsallis-Havrda-Charvat entropy is obtained from two-dimensional histogram which has determined by using the gray value of the pixels and the local average gray value of the pixels. Based on Tsallis-Havrda-Charvat entropy, we obtain the optimal threshold pair by maximizing the criterion function. The threshold pair groups the projection drawing of the 2-D histogram into four quadrants. Then we draw a line passing the optimal point. According to the line, we use the improved thresholding output function to separate the four quadrants into two parts, above the line and below the line. Therefore, the pixels are also grouped into two groups, targets and background. Experiment results show that the proposed method is robust to noise.
Adaptive background model
Xiaochun Lu, Yijun Xiao, Zhi Chai, et al.
An adaptive background model aiming at outdoor vehicle detection is presented in this paper. This model is an improved model of PICA (pixel intensity classification algorithm), it classifies pixels into K-distributions by color similarity, and then a hypothesis that the background pixel color appears in image sequence with a high frequency is used to evaluate all the distributions to determine which presents the current background color. As experiments show, the model presented in this paper is a robust, adaptive and flexible model, which can deal with situations like camera motions, lighting changes and so on.
MHT algorithm for distributed sensors networking
HongFei Wang, Qian Zhu
For the development cost, there is hardly any research on multi-hypotheses-tracking (MHT) for decentralized sensors networking system, although the distributed spot-level fusion for sensors networking is a key function in multi-target multi-sensor tracking system. MHT is a famous algorithm in target tracking, which is capable of track initiating, updating, deleting, keeping. This paper puts forward an MHT algorithm for decentralized sensors networking at spot fusion level which is validated by simulation results. Some important aspects in application are also put forward.
Multispectral Image Acquisition
icon_mobile_dropdown
Error analysis on laser measurement device of airborne LIDAR
Jianwei Wu, Hongchao Ma
The error sources of airborne laser measurement device which mainly include laser beam misalignment with respect to scanning mirror, clock error, scanner error and scanner torsion are discussed, and their effects to airborne LIDAR(Light Detection and Ranging)position accuracy are analyzed. Specially, laser beam misalignment's influences to LIDAR scanning line distortion and positioning accuracy are analyzed in detail quantitatively and qualitatively for oscillating scanner. The analysis demonstrates that laser beam misalignment influences the scanning line distortion and positioning accuracy more and more with the increasing height and scanning angle and can't be eliminated by common calibration methods but by the calibration method in factory for it is related to other error sources.
High resolution range profile imaging of high speed moving targets based on fractional Fourier transform
Min Cao, Yaowen Fu, Weidong Jiang, et al.
Due to high speed moving of the target, the wideband radar echoes after dechirping is a multi-component LFM signal. Directly using the fast Fourier transform (FFT) to implement pulse compression, the high resolution range profile (HRRP) will be broadened and distorted. To improve the quality of the HRRP, considering the fractional Fourier transform (FRFT) is the generalization of the FFT with good concentration for the LFM signal and without cross-term interference, we substitute the optimal FRFT for the FFT in the conventional HRRP imaging method and propose a new imaging method. For the proposed method, selecting the optimal order of the FRFT is important and a fast search algorithm based on fractional autocorrelation is employed in this paper. Compared with other HRRP imaging methods with the motion compensation, the proposed method not only is easy to perform without involvement of complex motion compensation, but also can correct the phase distortion, which is important for the following ISAR imaging process. Furthermore the method has low computational cost and good robust property for additive noises. Experimental results show the effectiveness of the proposed method.
Research of detection range of LWIR dispersive imaging spectrometer
Hong Xu, Xiangjun Wang
Among various types of long-wave infrared spectrometers, dispersive spectrometer has the simplist structure and has shown good service behavior in all kinds of imaging spectroscopy applications. For the technology of infrared remote imaging spectroscopy, an excellent designed imaging spectrometer is the key factor. Of all the designing factors, the tradeoff between the detection performance and the structure size is always the main topic. In this paper, according to the performance parameters of the used infrared sensor, characteristic of energy transmission of dispersive spectrometer, characteristic of target radiation and the effect of atmospheric transmission, a comprehensive research is done to construct a perfect receiving radiation model of LWIR dispersive spectrometer, providing theoretical foundations for accurately determining the relationship between instrumental structure parameters and the detection range. A designing model with an uncooled microbolometer as its detector is calculateded to analyse its potential in the application of long-wave remote imaging spectroscopy.
Real-time detecting system for infrared small target
Zhenjun Zhang, Zhiguo Cao, Tianxu Zhang, et al.
The paper presents a design method for high-speed image processing system based on FPGA+DSP+ASIC structure. According to the feature of the devices, the system reconfigures the data path and distributes the algorithm flexibly among DSP, FPGA, and the ASIC. Based on the above processing elements, an infrared small target detecting parallel system is constituted, which makes better use of the advantages of DSP, FPGA and ASIC while avoiding their disadvantages. Finally, the prototype test shows good performance and can meet the expected design target related to real-time processing.
Numerical simulation of SAR raw signal for ocean wave
Ding Guo, Xingfa Gu, Tao Yu, et al.
In order to study the SAR image of the ocean wave and the wave modulation of the RCS (radar cross section), the numerical simulation has been done by using the velocity bunching (VB) model with various parameters. The SAR image has a migration in the azimuthal direction for the wave velocity and the process of the nonlinear imaging. The way is an available method of studying the ocean wave image by SAR.
Vision bionics and application on the design of imaging guidance head
Meibo Lv, Fang Zi, Yanjun Li
Information given by a single-waveband imaging sensor doesn't satisfy the battlefield's needs. Many excellent characteristics of biotic vision are gradually applied in the design of intelligent imaging-guided missile. Combined with the function of lateral inhibition network in vision, a schematic diagram of intelligent infrared imaging-guided guidance head is proposed. Enlightened by the large field (LF) and small field (SF) of fly, the paralleled implement scheme of spatial double mode is proposed. A physical model of infrared imaging guidance head is given. The guidance head simulates the whole imaging progress of fly's ommateum. And its field of view is 360°. Then the synthetical application flow on imaging guidance referenced fly's vision system is proposed. The exploring study is beneficial and referenced to the design of intending imaging-guided system.
Research for imaging of high resolution airborne SAR data based on parameter estimation
Shuyuan Chao, Baixiao Chen, Peng Zhou
This paper proposes an algorithm combining classic CS algorithm with Dechirp technique for the process of imaging with big range cell migrations and partial aperture data. A great deal of interpolation will be needed if the azimuth pulse compression is accomplished by using a matched filter with only partial aperture data. So the CS algorithm is used to correct the range cell migration and complete the range pulse compression, while the Dechirp technique to accomplish the azimuth pulse compression. In addition, motion compensation is added because of the instability of the airborne SAR that leads to a debasement of the image. A curve of displacements in the LOS direction and phases of azimuth errors can be fitted by estimating the data, Thus it is easy to compensate them.
IR image generation of space target based on OpenGL
Tongsheng Shen, Ming Guo, Chenggang Wang
IR Scene simulation has been an important way to design and assess the IR sensor, and the key of simulation is the generation of IR scene image. Based on OpenGL, the method of IR image generation is proposed. The geometry model is constructed with professional CAD software, and the observer location is determined after scene transformation. The full infrared model of space target is built based on infrared physics and heat transfer, which includes the radiation, convection, conduction between different parts of the space target, and which also includes the radiation, convection of environment. Radiance of space target is converted to gray value, and properties of scene are defined according to the gray level. After a series of processing, dynamic IR images are generated with the technology of double buffering.
Effects of currents and winds on shallow water bathymetry SAR images
Kaiguo Fan, Weigen Huang, Bin Fu, et al.
Analytical expression of normalized radar backscattering cross-section (NRCS) contrasts for shallow water bathymetry SAR images has been firstly obtained by applying the continuity equation and the first order Bragg backscattering theory, and replacing the surface wave action balance equation with high frequency ocean wave spectrum balance equation. Based on the expression, we have simulated C band NRCS contrasts under different currents and sea surface winds firstly, results show the trend of NRCS contrasts' change with the shallow water bathymetry, and preferable currents and sea surface winds conditions for mapping shallow water bathymetry by SAR, which agree well with observations and shallow water bathymetry SAR images simulations. Moreover, the NRCS contrasts index which means the relationship between each currents and sea surface winds under given bathymetry parameters is first quantitatively calculated, which gives scientific proof for carrying SAR shallow water bathymetry surveys according to the ocean and weather conditions.
Modeling and simulation of CO2 laser initiative imaging radar system
Haiyan Li, Youjin He, Min Zhu
In all of precise guidance technology, imaging laser radar guidance technology is one of the most important orientations in the future development of precision guidance. The laser radar is a radar system adopting laser photosource and operating on optical band and the laser heterodyne measurement is a detection technique with high sensitivity. It can advance the hitting precision, the ability of resisting disturbing and breaching defence greatly. The numerical simulation technology can simulate some results of outfield experiments to a certain extent and becomes a new method for imaging laser radar guidance system design. This paper simulates the imaging laser radar which adopts CO2 laser as photosource and heterodyne detection. Our laboratory established the CO2 laser coherent imaging radar system. It consists of CO2 laser, acousto-optic modulators, beam-expending telescope, magnifying circuits and control computer etc. Firstly, the ranging equation of laser radar is analyzed. Then five models for the simulations of laser imaging radar are established including atmospheric transmission model, laser transmitter and laser receiver model, noise model, targets model, image processing and target recognizing model. The stimulation system of CO2 laser initiative imaging radar can operate easily and foresee the questions of the practicality design. The system had finished the elementary simulation and could evaluate accurately the capability of laser radar through the system's numerical simulation. It can estimate the key characteristic parameters and help to project design and capability advancing for experimental system.
Numerical analysis of aero-optic effects induced by the turbulence field surrounding hypersonic aicrafts
Wenxia Yang, Chao Cai, Mingyue Ding, et al.
The goal of this paper is to numerically simulate and analyze the aero-optic effects caused by the hyper-speed turbulence fields surrounding the aircraft under different flight conditions, and to characterize them with the associated optical transfer functions. First, analysis and computation of the aero-optic effects under different flight conditions have been addressed, where the parameters characterizing the hyper-speed turbulence field were obtained by solving its N-S equations via CFD methods. The infrared ray trajectories passing through the flow field with a non-homogeneous distribution of the refraction indices were acquired using the gradient index ray-tracing method, and the transfer function to represent the aero-optic effects was derived considering the principles of Fourier optics. The simulation results showed that the aero-optic transfer function is characterized as a low-pass filter of nonlinearly varying phases, which results the blurring and shifting of the objects in the acquired images.
Novel all-reflective Fourier transform imaging spectrometer based on Fresnel double-mirror
Dan Zhang, Ningfang Liao, Minyong Liang, et al.
In order to develop new technology during the spaceborne remote sensing mission, we introduce a novel all-reflective Fourier Transform Imaging Spectrometer(FTIS) based on Fresenel Double Mirror (FDM). Using all reflective parts, the system can work in visible and infrared waveband. Besides, the novel optical configuration which is involved can lead to the characters of the high spatial, spectral resolution and large field of view. The basic performance of the system was analyzed based on the test results.
Simultaneous acquisition of hyper-spectral image using the computed tomography imaging interferometer
Yu Lin, Ningfang Liao, Xinquan Wang, et al.
In this paper, we introduce a new approach of simultaneous acquisition of hyper-spectral image by means of using Computed Tomography Imaging Interferometer. Detecting rapidly varying targets both spatially and spectrally using imaging spectrometer is an international phenomenon in remote sensing in recent years and worthwhile in many domains such as pollution inspection and biochemical arms detection. Computed Tomography Imaging Interferometer (CTII) we put forward previously is a kind of imaging spectrometer with merits of high resolution, high throughout and high signal-to-noise rate. In our current work, some transmogrifications have been brought into the existed model of CTII. More channels are constructed in the new type of CTII-Simultaneous CTII (SiCTII). The rotation system in each channel is made irrotianal. Snapshot can be brought in the acquisition by means of opening and closing channels simultaneously. The Algebraic Reconstruction Technique (ART) and Artificial Neural Network (ANN) are considered in the image reconstruction. Referring to the characteristics of SiCTII, a conclusion can be made that SiCTII is a kind of imaging spectrometer for rapid acquisition and can be applied in the instances, in which the spatial resolution is not highly demanded and rapid data acquisition is demanded.
Wind retrieval with an assimilation method
Ming Wei, Huaying Yu, Liping Tang, et al.
In the study of radar wind field retrieval with variational assimilation method, the simple adjoint models have merits of simplifying calculation, saving time and improving efficiency. This paper studies wind field retrieval on levels of different height using the reflectivity advection equation as the control equation, proposes the parameterization method of sub-grid scale term which is related to the vertical motion. The flow pattern comparison and precision analysis with dual-Doppler radar have been made. Experiment results show this model could fully use the evolution information of reflectivity and effectively approximate to the real wind field.
Feature point detection from point cloud based on repeatability rate and local entropy
Jianjie Wu, Qifu Wang
An algorithm to detect feature points directly from unorganized point set is proposed. The algorithm introduces local entropy change of data points on local neighbors as a detection criterion to classify points according to the likelihood that they belong to a feature by making use of the characteristic that local entropy changes sharply in regions where surface changes great. Repeatability rate is introduced as well to reflect the frequency that a sample is detected as a feature point at different local windows. Size of the local neighborhoods is used as a discrete scale parameter to control size of the feature details. Experiment results show that the multi-scale feature point detection can improve the reliability of the detection phase and makes the algorithm more robust in the presence of noise. Furthermore, non-uniformly sampled point cloud can be dealt with.
Passive microwave imaging by aperture synthesis technology
Liang Lang, Zuyin Zhang, Wei Guo, et al.
In order to verify the theory of aperture synthesis at low expense, two-channel ka-band correlation radiometer which is basic part of synthetic aperture radiometer is designed firstly before developing the multi-channel synthetic aperture radiometer. The performance of two-channel correlation radiometer such as stability and coherence of visibility phase are tested in the digital correlation experiment. Subsequently all required baselines are acquired by moving the antenna pair sequentially, corresponding samples of the visibility function are measured and the image of noise source is constructed using an inverse Fourier transformation.
Optical registration analysis of multi-spectral airborne camera with multi-lens
Junyong Fang, Xiaoxue Dai, Bing Zhang, et al.
Optical registration requires the same spatial coordinate of all the corresponding pixels in the multi-spectrum images. The imaging equation of the multi-lens camera is established in the same ground coordinate. Some factors are analyzed and simulated with the actual parameters of a multi-spectral camera. The ground coordinate differences are compared among these factors, and the results are shown in paper. The shutter delay between the camera shutter and POS should be considered firstly, and the influence by focal length is more marked than the influence by angle of optical axis. The multi-spectral cameras should be selected to keep higher shutter synchronism.
Monopulse radar 3-D imaging and application in terminal guidance radar
Hui Xu, Guodong Qin, Lina Zhang
Monopulse radar 3-D imaging integrates ISAR, monopulse angle measurement and 3-D imaging processing to obtain the 3-D image which can reflect the real size of a target, which means any two of the three measurement parameters, namely azimuth difference beam elevation difference beam and radial range, can be used to form 3-D image of 3-D object. The basic principles of Monopulse radar 3-D imaging are briefly introduced, the effect of target carriage changes(including yaw, pitch, roll and movement of target itself) on 3-D imaging and 3-D moving compensation based on the chirp rate μ and Doppler frequency f d are analyzed, and the application of monopulse radar 3-D imaging to terminal guidance radars is forecasted. The computer simulation results show that monopulse radar 3-D imaging has apparent advantages in distinguishing a target from overside interference and precise assault on vital part of a target, and has great importance in terminal guidance radars.
Aerosol polarimetry sensor for the Glory Mission
Richard J. Peralta, Carl Nardell, Brian Cairns, et al.
This paper describes the Glory Mission Aerosol Polarimetry Sensor (APS) being built by Raytheon under contract to NASA's Goddard Space Flight Center. Scheduled for launch in late 2008, the instrument is part of the US Climate Change Research Initiative to determine the global distribution of aerosols and clouds with sufficient accuracy and coverage to establish the aerosol effects on global climate change as well as begin a precise long-term aerosol record. The Glory APS is a polarimeter with nine solar reflectance spectral bands that measure the first three Stokes parameters vector components for a total of 27 unique measurements. In order to improve the reliability and accuracy of the measurements, additional 9 redundant measurements are made, yielding a total of 36 channels. The sensor is designed to acquire spatial, temporal, and spectral measurements simultaneously to minimize instrumental effects and provide extremely accurate Raw Data Records. The APS scans in the direction close to of the spacecraft velocity vector in order to acquire multi-angle samples for each retrieval location so that the Stokes parameters can be measured as functions of view angle.