Proceedings Volume 6977

Optical Pattern Recognition XIX

David P. Casasent, Tien-Hsin Chao
cover
Proceedings Volume 6977

Optical Pattern Recognition XIX

David P. Casasent, Tien-Hsin Chao
View the digital version of this volume at SPIE Digital Libarary.

Volume Details

Date Published: 16 March 2008
Contents: 9 Sessions, 26 Papers, 0 Presentations
Conference: SPIE Defense and Security Symposium 2008
Volume Number: 6977

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
  • Front Matter: Volume 6977
  • Invited Papers I
  • Invited Papers II
  • Pattern Recognition Correlators
  • Pattern Recognition Filters and Applications
  • Image Processing
  • Tracking and Applications
  • Poster Session: Pattern Recognition Filters and Applications
  • Poster Session: Image Processing
Front Matter: Volume 6977
icon_mobile_dropdown
Front Matter: Volume 6977
This PDF file contains the front matter associated with SPIE Proceedings Volume 6977, including the Title Page, Copyright information, Table of Contents, Introduction (if any), and the Conference Committee listing.
Invited Papers I
icon_mobile_dropdown
Optical ID tags for automatic vehicle identification and authentication
We review the potential of optical techniques in security tasks and propose to combine some of them in the design of new optical ID tags for automatic vehicle identification and authentication. More specifically, we propose to combine visible and near infrared imaging, optical decryption, distortion-invariant ID tags, optoelectronic devices, coherent image processor, optical correlation, and multiple authenticators. A variety of images and signatures, including biometric and random sequences, can be combined in an optical ID tag for multifactor identification. Encryption of the information codified in the ID tag allows increasing security and deters from unauthorized usage of optical tags. A novel NIR ID tag is designed and built by using commonly available materials. The ID tag content cannot be visually perceived at naked eye; it cannot be copied, scanned, or captured by any conventional device. The identification process encompasses several steps such as detection, information decoding and verification which are all detailed in this work. Design of rotation and scale invariant ID tags is taken into account to achieve a correct authentication even if the ID tag is captured in different positions.
Distortion-invariant kernel filters for general pattern recognition
Rohit Patnaik, David Casasent
We note several key general pattern recognition (GPR) issues that have been ignored in all prior distortion-invariant kernel filter (kernel DIF) work. These include: the unrealistic assumption of centered test data, the lack of a fast FFTbased on-line implementation, the significantly larger storage and on-line computation requirements, incorrect formulation of the kernel filter in the FT domain, incorrect formulation of prior image-domain kernel SDF and Mace filters, and the unrealistic use of test set data for parameter selection. We present several improvements to prior kernel DIF work. Our primary objective is to examine the viability of kernel DIFs for GPR and automatic target recognition (ATR) applications (where the location of the object in the test input is not known). Thus, in this paper, we apply our improved kernel DIFs to CAD ATR data. We address range and full 360° aspect view variations; we also address rejection of unseen confuser objects and clutter. We use training and validation set data (not test set data) to select the kernel parameter. We show that kernel filters (higher-order features) can improve classification and confuser rejection performance. We consider only kernel SDF filters, since their on-line computation requirements are reasonable; we present test results for both polynomial and Gaussian kernels. The main purposes of this paper are to: note issues of importance ignored in all prior kernel DIF work, detail how to properly perform energy minimization in kernel DIFs, show that kernel SDF filters can correct errors for ATR data, and compare the performance of kernel SDF filters and standard Minace DIFs. We also introduce our new Minace-preprocessed kernel SDF filter.
Grayscale optical correlator for CAD/CAC applications
This paper describes JPL's recent work on high-performance automatic target recognition (ATR) processor consisting of a Grayscale Optical Correlator (GOC) and neural network for various Computer Aided Detection and Computer Aided Classification (CAD/CAC) applications. A simulation study for sonar mine and mine-like target detection and classification is presented. Applications to periscope video ATR is also presented.
Invited Papers II
icon_mobile_dropdown
Multiple target detection in video using quadratic multi-frame correlation filtering
Most integrated target detection and tracking systems employ state-space models to keep track of an explicit number of individual targets. Recently, a non-state-space framework was developed for enhancing target detection in video by applying probabilistic motion models to the soft information in correlation outputs before thresholding. This framework has been referred to as multi-frame correlation filtering (MFCF), and because it avoids the use of state-space models and the formation of explicit tracks, the framework is well-suited for handling scenes with unknown numbers of targets at unknown positions. In this paper, we propose to use quadratic correlation filters (QCFs) in the MFCF framework for robust target detection. We test our detection algorithm on real and synthesized single-target and multi-target video sequences. Simulation results show that MFCF can significantly reduce (to zero in the best case) the false alarm rates of QCFs at detection rates above 95% in the presence of large amounts of uncorrelated noise. We also show that MFCF is more adept at rejecting those false peaks due to uncorrelated noise rather than those due to clutter and compression noise; consequently, we show that filters used in the framework should be made to favor clutter rejection over noise tolerance.
Dynamic range compression deconvolution for enhancement of automatic target recognition system performance
Bahareh Haji-saeed, Jed Khoury, W. D. Goodhue, et al.
A generic nonlinear dynamic range compression deconvolver (DRCD) is proposed. We have performed the dynamic range compression deconvolution using three forms of nonlinearities: (a) digital implementation- A-law/μ-law, (b) hybrid digital-optical implementation- two-beam coupling photorefractive holography, and (c) all optical implementation- MEMS deformable mirrors. The performance of image restoration improves as the saturation nonlinearity increases. The DRCD could be used as a preprocessor for enhancing Automatic Target Recognition (ATR) system performance. In imaging through atmosphere, factors such as rain, snow, haze, pollution, etc. affect the received information from a target; therefore the need for correcting these captured images before an ATR system is required. The DRCD outperforms well-established image restoration filters such as the inverse and the Wiener filters.
Mine detection in multispectral imagery data using constrained energy minimization
M. I. Elbakary, M. S. Alam
Multispectral imagery is used for a wide variety of military and commercial applications, including object detection such as mines. The main reason for using multispectral imagery is that it reveals spectral information about the scene which cannot be obtained from a single spectral band. This paper introduces a new algorithm for mine detection in multispecral imagery using the constrained energy minimization (CEM) approach. The CEM approach is introduced as classifier. The novelty of this idea is that this classifier uses only the information of the mines for training and enabling the potential mines without using information about the clutter in the scene. Using only mines information for detection is a major advantage of the CEM approach. In addition, the CEM approach is modified such that recomputing the autocorrelation matrix is not necessary and using the algorithm became scene independent Then, to reduce the false alarm further, morphological processing and stochastic expectation maximization (SEM) algorithm are employed for post-processing. The results of the proposed algorithm were promising when the algorithm is tested using real multispectral imagery.
Pattern Recognition Correlators
icon_mobile_dropdown
M-ary pseudorandom phase masks for spatially efficient phase encoded reference JTC
In this paper we propose the use of discrete pseudorandom phase mask for a spatially efficient phase encoded JTC. In the proposed JTC system the reference image is phase encoded using a pseudorandom phase mask to eliminate extraneous peaks cluttering or overshadowing the correlation output. The phase encoding scheme also eliminates the need for a spatial separation in the joint input image resulting in the full use of the SLM. An optoelectronic architecture of the proposed system is presented. To relax the need for a phase-only SLM with a high phase resolution we proposed the use of discrete pentary pseudorandom phase mask.
Probability density function-based Fisher ratio applied to polarization-enhanced patterns
Aed El-Saba, M. S. Alam, H. Nalluri
The use of Fisher ratio (FR) algorithm to predict a pattern in an input seen has been applied in several applications in the literature with different success rate, depending on how close is the similarity of the statistical parameters between the background and the patterns. We propose a modification to the FR ratio algorithm that is dependent on the probability density function (PDF). The modified PDF-FR algorithm provides good improvements over that of the PDF used alone. We further enhance the performance of the PDF-FR using polarization-enhanced imagery.
Pattern recognition using Gaussian-filtered, shifted phase-encoded fringe-adjusted joint transform correlation
Pattern recognition for real-time applications requires the detection scheme be a simple architecture, fast in operation, able to detect all the potential targets without generating any false alarms, and invariant to noise and distortion. Though several target detection algorithms have been proposed in the literature over the years, but most of them are found to be not as efficient in meeting all the above-mentioned objective requirements. A new Gaussian-filtered, shifted phase-encoded fringe-adjusted joint transform correlation technique has been developed in this paper for an optical pattern recognition system. The input noisy image is first filtered by using a Gaussian filter, which helps in overcoming the effect of background noise and distortions. Then the filtered image is correlated with the reference image using the proposed joint transform correlator, which eliminates the problems of duplicate correlation heights, false alarms and low discrimination ratio. The architecture involves optical devices including lenses and spatial light modulators, which guarantees the very fast operation required for real-time applications. Computer simulation results show that the algorithm can successfully discriminate between targets and non-targets contained in the input scene even in the presence of noise and can also make the best utilization of the correlation space.
LPCC invariant correlation filters: realization in 4-f holographic correlator
N. N. Evtikhiev, S. N. Starikov, S. A. Sirotkin, et al.
This paper contains the results of synthesis and realization of linear phase coefficient composite (LPCC) filters in 4-f correlator. LPCC filters application allows achieving invariance of correlation peak in the presence of geometric distortions of contour objects. LPCC filters were realized as computer generated binary amplitude holograms for application in optical correlator. Experimental results of invariant pattern recognition are presented.
Pattern Recognition Filters and Applications
icon_mobile_dropdown
Space vehicle pose estimation via optical correlation and nonlinear estimation
John M. Rakoczy, Kenneth A. Herren
A technique for 6-degree-of-freedom (6DOF) pose estimation of space vehicles is being developed. This technique draws upon recent developments in implementing optical correlation measurements in a nonlinear estimator, which relates the optical correlation measurements to the pose states (orientation and position). For the optical correlator, the use of both conjugate filters and binary, phase-only filters in the design of synthetic discriminant function (SDF) filters is explored. A static neural network is trained a priori and used as the nonlinear estimator. New commercial animation and image rendering software is exploited to design the SDF filters and to generate a large filter set with which to train the neural network. The technique is applied to pose estimation for rendezvous and docking of free-flying spacecraft and to terrestrial surface mobility systems for NASA's Vision for Space Exploration. Quantitative pose estimation performance will be reported. Advantages and disadvantages of the implementation of this technique are discussed.
Improved training for target detection using Fukunaga-Koontz transform and distance classifier correlation filter
M. I. Elbakary, M. S. Alam, M. S. Aslan
In a FLIR image sequence, a target may disappear permanently or may reappear after some frames and crucial information such as direction, position and size related to the target are lost. If the target reappears at a later frame, it may not be tracked again because the 3D orientation, size and location of the target might be changed. To obtain information about the target before disappearing and to detect the target after reappearing, distance classifier correlation filter (DCCF) is trained manualy by selecting a number of chips randomly. This paper introduces a novel idea to eliminates the manual intervention in training phase of DCCF. Instead of selecting the training chips manually and selecting the number of the training chips randomly, we adopted the K-means algorithm to cluster the training frames and based on the number of clusters we select the training chips such that a training chip for each cluster. To detect and track the target after reappearing in the field-ofview ,TBF and DCCF are employed. The contduced experiemnts using real FLIR sequences show results similar to the traditional agorithm but eleminating the manual intervention is the advantage of the proposed algorithm.
Multiscale beamlet transform application to airfield runway detection
The context-driven target recognition requires the object-of-interest (OOI) to be first detected. We use the multiscale beamlet transform to detect airport runways as the OOI for detecting the aircraft. The up-to-down strategy in the beamlet graph structure is used for the connectivity and directional continuation of the edges, which are first detected in a coarse scale and are then refined at several finer scales.
Land cover mapping after the tsunami event over Nanggroe Aceh Darussalam (NAD) province, Indonesia
H. S. Lim, M. Z. MatJafri, K. Abdullah, et al.
Remote sensing offers an important means of detecting and analyzing temporal changes occurring in our landscape. This research used remote sensing to quantify land use/land cover changes at the Nanggroe Aceh Darussalam (Nad) province, Indonesia on a regional scale. The objective of this paper is to assess the changed produced from the analysis of Landsat TM data. A Landsat TM image was used to develop land cover classification map for the 27 March 2005. Four supervised classifications techniques (Maximum Likelihood, Minimum Distance-to- Mean, Parallelepiped and Parallelepiped with Maximum Likelihood Classifier Tiebreaker classifier) were performed to the satellite image. Training sites and accuracy assessment were needed for supervised classification techniques. The training sites were established using polygons based on the colour image. High detection accuracy (>80%) and overall Kappa (>0.80) were achieved by the Parallelepiped with Maximum Likelihood Classifier Tiebreaker classifier in this study. This preliminary study has produced a promising result. This indicates that land cover mapping can be carried out using remote sensing classification method of the satellite digital imagery.
Image Processing
icon_mobile_dropdown
Multifractal and directional wavelet analysis for accurate detection of precipitation events in weather radar images
In this paper, we propose the use of directional Gabor filtering and multifractal analysis based quality control (QC) to provide accurate identification of precipitation in weather data collected from meteorological-radar volume scans. The QC algorithm is an objective algorithm that minimizes human interaction. The algorithm utilizes both textural and intensity information obtained from the two lower-elevation reflectivity maps. Computer simulations are provided to show the effectiveness of this algorithm.
Noise elimination methods in topological pattern recognition
As we published in the last three years, we can use the continuity threshold and the geometric projection methods to eliminate the background noise and the cloud-obscuring noise in an edge-detected 2D color picture for the topological pattern recognition system developed by the author. Preliminary computer experiments showed that the background noise elimination is thorough and the reconstruction of the obscured part of the original image is 90%+ accurate.
Image watermarking extraction using Fourier domain Wiener filter
Philip Birch, Marios Pavlidis, Ankit Panwar, et al.
Digital watermarking is a vital process for protecting the copyright of images. This paper presents a method of embedding a private robust watermark into a digital image. The full complex form the Wiener filter is used to extract the signal from the watermarked image. This is shown to outperform the more conventional approximate notation. The results are shown to be extremely noise insensitive.
Computational sensing algorithms for image reconstruction and the detection of moving objects in multiplexed imaging systems
The problem of wide area persistent surveillance presents imaging problems which cannot be addressed by traditional sensing. We consider a coded aperture approach to imaging a wide area with high resolution for an object tracking application. Coded aperture imaging systems are generally designed for obtaining images of static scenes. For exploitation of dynamic scenes, the coding approach must be modified to not only reconstruct the image, but also to facilitate the detection of moving objects over this large area. We present a multi-scale framework that describes a multiplexed sensing and image reconstruction process. A novel method is introduced for learning a "motion model" for a given scene, and using it to handle the ambiguity induced by object motion. The results of initial simulations are presented.
Tracking and Applications
icon_mobile_dropdown
Near real-time extraction of planar features from 3D flash-ladar video frames
This paper describes a novel method used to extract planar surfaces from a stream of 3D images in near real-time. The method currently operates on 3D images acquired from a MESA SwissRanger SR-3000 infrared time of flight camera, which operates in a manner similar to flash-ladar sensors; the camera provides the user with range and intensity value for each pixel in the 176 by 144 image frame. After application of the camera calibration the range measurement associated with each pixel can be converted to a Cartesian coordinate. First, the proposed method splits the focal image plane into sub-images or sub-windows. The method then operates in the 3D parameter space to find an estimate of the planar equation best describing the point cloud associated with the window pixels and to compute a metric that defines how well the sub-window points fit to the planar estimate. The best fit sub-window is then used as an initialization to one of two investigated methods: a parameter based search technique and cluster validation using histogram thresholding to extract the entire plane from the 3D image frame. Once a plane is extracted, a feature vector describing that plane along with their describing statistics can be generated. These feature vectors can then be used to enable feature-based navigation. The paper will fully describe the feature extraction method and will provide application results of this method to extract features from indoor 3D video data obtained with the MESA SwissRanger SR-3000. Also provided is a brief overview of the generation of feature statistics and their importance.
Compact liquid crystal waveguide based Fourier transform spectrometer for in-situ and remote gas and chemical sensing
Vescent Photonics Inc. and Jet Propulsion Lab are jointly developing an innovative ultra-compact (volume < 10 cm3), ultra-low power (<10-3 Watt-hours per measurement and zero power consumption when not measuring), completely non-mechanical electro-optic Fourier transform spectrometers (EO-FTS) that will be suitable for a variety of remoteplatform, in-situ measurements. These devices are made possible by a novel electro-evanescent waveguide architecture, enabling "chip-scale" EO-FTS sensors. The potential performance of these EO-FTS sensors include: i) a spectral range throughout 0.4-5 μm (25000 - 2000 cm-1), ii) high-resolution (▵λ≤ 0.1 nm), iii) high-speed (< 1 ms) measurements, and iv) rugged integrated optical construction. This performance potential enables the detection and quantification of a large number of different atmospheric gases simultaneously in the same air mass and the rugged construction will enable deployment on previously inaccessible platforms. The sensor construction is also amenable for analyzing aqueous samples on remote floating or submerged platforms. To date a proof-of-principle prototype EO-FTS sensor has been demonstrated in the near-IR (range of 1450-1700 nm) with a 5 nm resolution. This performance is in good agreement with theoretical models, which are being used to design and build the next generation of EO-FTS devices.
Predictive control and resource management of a distributed coastal monitoring sensor network
Ashit Talukder, Anand Panangadan
A novel model predictive control (MPC) technique is used as a general framework for resource management in sensor networks. The MPC formulation adapts the sensor network system parameters that impact the energy consumption rate (such as sensor sampling rates) to variations in the criticality of the phenomenon being monitored. This approach is illustrated using two examples. The first is based on a sensor network where the data is temporal in nature. The second is based on an coastal environment monitoring network where the data is spatiotemporal in nature and event criticality shows variation in both space and time. Simulation results from both these applications are presented that demonstrate the functioning of the proposed predictive controller in sensor network control.
Poster Session: Pattern Recognition Filters and Applications
icon_mobile_dropdown
Using commercial photo camera’s RAW-based images in optical-digital correlator for pattern recognition
In optical-digital correlators for pattern recognition, linear registration of correlation signals is significant for both of recognition reliability and possible input image restoration. This usually achieves with scientific graduated technical cameras, but most of commercial digital cameras now have an option of RAW data output. With appropriate software and parameters of processing, it is possible to get linearized image data from photo camera's RAW file. Application of such photo cameras makes optical-digital systems cheaper, more flexible and brings along their wider propagation. For linear registration of correlation signals, open-source Dave Coffins's RAW converter DCRAW was used in this work. Data from photo camera were linearized by DCRAW converter in "totally RAW documental mode" with 16-bit output. Experimental results of comparison between linearized and non-linearized correlation signals and digitally restored input scene images are presented. It is shown, that applied linearization allows to increase linear dynamic range for used Canon EOS 400D camera more that 3 times.
Poster Session: Image Processing
icon_mobile_dropdown
High-spatial resolution land cover mapping using remotely sensed image
We attempted to investigate the potential of using satellite image for acquiring data for remote sensing application. This study investigated the potential of using digital satellite image for land cover mapping over AlQasim, Saudi Arabia. Satellite digital imagery has proved to be an effective tool for land cover studies. Supervised classification technique (Maximum Likelihood, ML, Minimum Distance-to- Mean, MDM, Parallelepiped, P) techniques were used in the classification analysis to extract the thematic information from the acquired scenes. Besides that, neutral network also performed in this study. The accuracy of each classification map produced was validated using the reference data sets consisting of a large number of samples collected per category. The study revealed that the ML classifier produced better result. The best supervised classifier was chosen based on the highest overall accuracy and Kappa statistic. The results produced by this study indicated that land cover features could be clearly identified and classified into a land cover map. This study suggested that the land cover types of AlQasim, Saudi Arabia can be accurately mapped.
The cognitive structural approach for image restoration
Igor Mardare, Veacheslav Perju, David Casasent
It is analyzed the important and actual problem of the defective images of scenes restoration. The proposed approach provides restoration of scenes by a system on the basis of human intelligence phenomena reproduction used for restoration-recognition of images. The cognitive models of the restoration process are elaborated. The models are realized by the intellectual processors constructed on the base of neural networks and associative memory using neural network simulator NNToolbox from MATLAB 7.0. The models provides restoration and semantic designing of images of scenes under defective images of the separate objects.
Face recognition on the basis of moment invariants, principal component analysis, and correlation
Veacheslav Perju, David Casasent, Igor Mardare, et al.
There are presented the results of investigation of the algorithms of invariant face recognition of masked persons. There are described 3 algorithms based on Image Moments Features, Principal Component Analyses algorithm and Correlation algorithm. It is presented the description of the elaborated software for PC based face recognition, created in Borland C++ Builder environment. There are presented the data of the face recognition in conditions of masking, change of the rotation, scale of the images.