This PDF file contains the front matter associated with SPIE Proceedings Volume 11438 including the Title Page, Copyright information, Table of Contents, Author and Conference Committee lists.

An optical and non-contact continuous measurement method to detect human blood pressure through a high-speed camera is discussed in this paper. With stable ambient light, photoplethysmographic (PPG) signals of face and palm area are obtained simultaneously from the video captured by high-speed camera, whose frame rate should be higher than 100 frames per second. Pulse transit time (PTT) is measured from the R-wave distance between the two PPG signals. The Partial least squares regression( PLSR) model was established to train the samples, and the relationship between PTT and blood pressure, including intra-arterial systolic pressure (SBP) and diastolic pressure (DBP), was established to obtain blood pressure. Compared with the output of traditional sphygmomanometer, the blood pressure data collected from non-contact system has little error and meets the fitting conditions. We first proposed an accurate video-based method for non-contact blood pressure measurement using machine learning, and the average error of SBP is 0.148mmHg and of DBP is 0.359mmHg.

A low-frequency photoacoustic (PA) sensor based on Parylene-C diaphragm for micro-leakage gas detection is presented. The overall structure of the sensor head includes a cylindrical brass shell, a Parylene-C diaphragm, and a PA cavity regarded as a Fabry-Perot (F-P) cavity as well. The volume of the PA cavity is only 74 μL. A distributed feedback (DFB) laser is used as an acoustic excitation source. The PA pressure signal is obtained by measuring the deflection of the Parylene-C diaphragm using fiber-optic white-light interference (WLI) method. The PA sensor head is used for acetylene (C₂H₂) detection. A detection limit of 36 parts-per-billion (ppb) is achieved when the lock-in integration time is 1 s.

In order to accurately measure the glucose concentration, the photoacoustic spectroscopy combined with the scanning position method was used in this paper. Meanwhile, a kind of custom-built photoacoustic detection system was established. In this system, a 532nm pumped OPO pulsed laser was used as the excitation source, and a non-focused ultrasonic detector was used to capture the photoacoustic signals of phantom and glucose. To improve the measurement accuracy of glucose by using the photoacoustic spectroscopy, a scanning sub-system was established to accurately locate the position of the simulated vessel. In the experiments, to simulate the bio-tissue and the blood vessel in human body, a specimen made of agar block with a carbon bar and a silicon gel pipe was prepared to verify the availability and feasibility of the established photoacoustic detection system. According to the obtained photoacoustic peak-to-peak values of the specimen by scanning method, the position of the carbon bar and the silicon gel pipe in the specimen can be located. Then, to test the photoacoustic detection of glucose in the silicon gel pipe, a cycling sub-system was used to simulate the blood flow in the blood vessel. Based on the located position of the silica gel pipe, the focused laser spot was shifted and accurately irradiated into the silica gel pipe to induce the photoacoustic signals of glucose. By mean of the photoacoustic system, several different concentrations of glucose solutions were test, and the time-resolved photoacoustic signals and peak-to-peak values of glucose are all obtained. The prediction model of glucose concentration was established by using the linear fitting method. At the same time, the predicted result was compared with that of the glucose in the agar specimen. Results show that the novel photoacoustic detection of glucose is available. The correction coefficient of the glucose concentration prediction was improved by using the photoacoustic spectroscopy combined with the scanning position method.

OH radicals in the upper and middle atmosphere are important oxidants and play an important role in atmospheric photochemistry. Current space-based payload can only obtain profile information. In view of the inhomogeneity of atmospheric two-dimensional distribution field, a hyper-spectral OH radical tomography detection scheme based on one-dimensional imaging spatial heterodyne spectroscopy is proposed. In this paper, the basic principle of SHS is introduced. Then it introduces the detection mode of computer chromatography and the principle of atmospheric inversion. Finally, it introduces the design of OH chromatography sounder based on the principle of SHS technology. In order to verify the effectiveness of the design, the data simulation is carried out.

The digital micromirror device DMD is widely used in visible light projection, special-purpose spatial light modulation, and infrared scene simulation, due to its high resolution, uniformity and energy concentration. In some applications that require high frame rate scene image display, it is necessary to ensure that the DMD displays high gray level images at a high frame rate. However, the display frame rate of the pulse width modulation (PWM) method is limited by the minimum time required for DMD loading data, unable to achieve high frame rate display. Although the DMD binary display mode can meet the requirements of high frame rate display, it cannot meet the requirements of high gray level. The paper proposes an image display method that uses light source and DMD to synchronize modulation. Decompose the high-order gray image into bit planes according to the gray threshold, and the DMD displays each bit plane in binary mode under the trigger of an sync pulse. The intensity of the illumination laser source is modulated by an acousto-optic modulator to match the bit plane and the radiated laser power. This method makes use of the function of high frame rate display in DMD binary mode cooperate with light intensity modulation of illumination laser source, and realizes high frame rate and high dynamic range image display. With this method, the maximum frame rate of 8-bit gray level image with 1920 ×1080 resolution can reach 2KHz. The experimental system has realized 200Hz frame rate display of 8-bit gray level image with 1920 ×1080 resolution.

The differences in material properties of various common recyclable garbage will be directly mapped to the difference in spectral characteristics. While acquiring the spectral information of the garbage, the hyperspectral imaging technology can obtain the spatial information of the garbage, and realize the rapid detection and classification of garbage by using the method of "spectral-spatial" .The classification model is established combined the spectral characteristics under the sample feature space with CNN (convolutional neural network) machine learning algorithm, and then the classification model is trained and optimized by using database training sample set. Finally, 92.10% classification accuracy is achieved after testing the sample set.

The color deviation, which refers to the different between the obtained image and the image under the standard light, is from the light source and reflection characteristic of the object. For satellite remote sensing, the length of the radiation transmission path is different for imaging of different latitudes, especially considering the effects of the atmosphere. Therefore, imaging for high latitude in the winter maybe brings some color deviation into the data. There have some color deviation detection methods for digital camera photo, but they are not fit for the remote sensing data, because of the large image range. In the work, a color deviation detection method for the satellite remote sensing image is developed. And the method is validated using the Landsat 8 images obtained in the winter and summer respectively.

Super-resolution reconstructed convolution neural network (SRCNN) is widely used in image quality improvement of single image. Traditional SRCNN training uses the loss function of minimum mean square error (MSE) and the method based on stochastic gradient descent (SGD) to optimize. Its learning rate adjustment strategy is limited by pre-specified adjustment rules, and it is difficult to select the initial value. Considering the complex texture and low resolution of remote sensing images, a deconvolution layer is proposed to replace the bi-cubic interpolation enlarged image in the traditional SRCNN network to overcome the mosaic effect. At the same time, Adam optimizer is used to control the network training. After considering the first and second moment estimation of gradient comprehensively, the update step is calculated. Thus, the adaptive update of learning rate is realized and the speed of network training is greatly accelerated. The simulation results show that this method has advantages in edge reconstruction and texture details compared with the conventional super-resolution reconstruction algorithm.

3D localization of point source is widely used in many fields, such as bioimaging and autonomous driving fields. However, the localization is hard to perform under scattering conditions because of the diffuse effect of the scattering. We propose a novel method for 3D localization of point source under scattering conditions based on light field imaging and deep learning by only one shot. First, we introduce the description of the point source in a light field wise and how to localize a point source by its light field. On the basis, we elaborate on the effect of scattering on a light field and how to retrieve the location of a point source from a light field with scattering. Then, the effect of aberration on a light field will be introduced. We also build an artificial scene and a deep learning framework to perform a 3D localization practically, and the feasibility and accuracy of our method have been evaluated.

Compound eyes of insect is an ideal miniaturized, multi-aperture and large-field-of-view optical system. It also has intelligent detection capability including high sensitivity for detecting moving targets, and high resolution for light intensity, wavelength (color) or polarization. In this paper, a compact visible bionic compound eyes system based on micro-surface fiber faceplate is studied, which uses nine micro-lens-groups of specific divergent angle lines of sight, and micro-surface fiber faceplate direct-coupled with large area (5120X5120 pixels) CMOS camera. The nine micro-lensgroups put the images onto micro-surfaces of fiber faceplate, implementing nine partially (~50%) overlapping field of views (FOVs) sub-aperture imaging system with real-time acquisition and output. Compound eyes system has features including the (stitching) large FOV, the super-resolution imaging of the overlapping FOVs, short-distance stereovision, and the intelligent high sensitivity for detecting moving targets. With polarizers or filters, it is a practical pattern can also implementing full-polarization imaging or multispectral imaging capability. In the fields of emergency obstacle avoidance, missile reconnaissance and short-range fuse, and underwater unmanned submarine navigation, compound eyes system has broad application prospects. At present, a compound eyes system with 80° FOV has been developed, and the real-time large-FOV image stitching method has been applied. In the system initialization stage, the image calibration and non-uniformity correction are pre-processed. In the real-time large-FOV image stitching stage, the CUDA parallel acceleration method is used. The single-frame stitching takes less than 30ms, which meets the requirement of real-time processing.

Noninvasive object imaging through strongly-scattering turbid layers has attracted the attention of many experts due to the potential application in the biomedical imaging and bioscience. The traditional speckle correlation method with Gerchberg-Saxton (GS) algorithm is used to restore object in a single-shot speckle pattern. However, this method suffers from the problems of convergence to local minimums, many iterations and cannot determine the object direction, due to the randomly assign initial value to GS algorithm. Bispectrum analysis method enables the directional Fourier phase retrieved using single-shot speckle pattern, but there are the problems of requiring high-resolution speckle pattern, low SNR and selecting the size of filter window. Therefore, we report an effective noninvasive imaging method through strongly-scattering turbid layers on the basis of bispectrum analysis and GS algorithm to restore the object in a lowresolution speckle pattern. Meanwhile, the new expression of Gaussian filter function is introduced contribute to determine the window size of filter in the processes of bispectrum analysis. In this proposed approach, the window size of filter is determined by the adjust factor according to the new form of Gaussian filter function, and the initial Fourier phase with directional information is generated by bispectrum analysis in a low-resolution speckle pattern. Then the initial Fourier phase is used as the randomly assign initial value to retrieve Fourier phase of object. Hence, the proposed method required no high-resolution, multiple iterations, nor randomly assign initial values to restore directional object. This work carries out simulations and experiments to demonstrates noninvasive object imaging in the low-resolution speckle pattern through strongly-scattering turbid layers.

Digital Image Correlation (DIC) is an optical measurement method that can measure the deformation and even the appearance of an object. Because of the superiority of non-contact, full field, and low environment requirements, DIC can be used to measure the high temperature mechanical properties of materials in extreme environment. However, there are still difficulties in applying DIC in high temperature displacement measurement, such as heat flow disturbance. It will distort the images acquired at high temperature and affect the measurement accuracy of DIC. In this paper, the disturbance characteristics caused by heat flow are studied and the corresponding correction method is proposed. Firstly, the causes of distortion and the distortion characteristics are analyzed and then verified by experiments. Second, a method for correcting the distortion caused by heat flow is proposed. Finally, the availability of the proposed correction method is verified by experiment.

In Computational Spectral imaging, two-dimensional coded apertures and dispersive elements realize the mixed modulation of spatial information and spectral information of the target respectively, and then reconstruct the threedimensional data cube. Therefore, coded aperture plays a vital role. In the imaging process, by moving the coded aperture to increase the number of measurements, the aperture moved one code element at each step to simulate the actual push-broom process. Three types of coded apertures were considered, which are Gauss random coded aperture, Hadamard coded aperture and Harmonic coded aperture, and the reconstruction effect of the three coded apertures were analyzed. The Least Square (LS) algorithm was considered to reconstruct three-dimensional data cube. Compared with the classical Two-step Iterative Shrinkage/Thresholding (TwIST) algorithm, the reconstructed Structural Similarity Index Measurement (SSIM) and Peak Signal to Noise Ratio (PSNR) by LS algorithm were better than TwIST algorithm. It was indicated that the SSIM and PSNR increased with the increasing number of measurements. When the number of measurements was similar with the number of spectral segments, the SSIM of the three coded apertures reached more than 0.9 by LS algorithm. However, the SSIM and PSNR of the Gauss random coded aperture were the largest Obviously, which are 0.995 and 52.560, respectively. And the PSNR of Gauss random coded aperture was 13 dB more than that of Hadamard and Harmonic coded apertures. When the number of measurements was constant, the SSIM and PSNR decrease gradually with the increasing number of spectral segments. The simulation results showed that the LS algorithm was superior to the TwIST algorithm in the reconstruction process, and the Gauss random coded aperture had the best performance.

3D imaging through a diffuser is a popular topic recently due to its extensive use, which is from earth observation to biomedical diagnosis. Although many methods have been proposed to realize this goal, such as holographic imaging and wavefront shaping, it is still complicated i.e. they require either a reference beam or sequential scanning to compensate the random phase. Here, we proposed a new verified approach without reference beam, just using a structured light illumination combined with time averaging technology to resolve the phase of the object. The result shows that this method is easy to get 3D image without speckle noise. Furthermore, it can also be used to detect the deformation of a hidden object and be extended to the biomedical imaging.

This study describes various methods for optical imaging through scattering media based on compression detection. In these methods, low-intensity micro-structured light patterns are launched sequentially onto an object by using a digital micro-mirror device. The corresponding light intensity information is collected and an intensity correlation algorithm is used to recover the information of the object in the scattering media. For each specific sampling reconstruction method, its related systems are listed and its parameter indicators and application scenarios are analyzed. Prospects for the development of single-pixel imaging through scattering media are also discussed.

The accuracy of star centroid extraction and identified star number is both crucial features for star sensor precision. Motion blur and fracture are introduced when star sensor works under high dynamic conditions, which affect the accuracy of star centroid extraction and further reduce the precision of the star sensor. To improve the precision of star sensor, this paper proposes an adaptive window star map restoration method based on energy equalization. The local degradation function is estimated for the star point energy distribution region, and dynamic star map simulation and restoration are performed. The simulation process is divided into three steps. Firstly, establish the motion trajectory of the star point centroid in the detector plane during the exposure time, according to the trailing trajectory size and direction adaptive selection window. Secondly, accumulate the star image point energy in the adaptive window to achieve star point recovery, and finally the recovery effect was evaluated by star extraction results.

In this paper, a dynamic detection system for thermocouple cable insulation defects is proposed. The common thermocouple cable defects are insulation cracking and cable joints. These two defects can be identified by detecting the cable diameter. Detecting such defects through the human eye is time consuming and labor intensive and does not guarantee quality. The method in this paper is to make the thermocouple line pass the field of view of the line camera when moving at high speed. Then use the edge detection algorithm to calculate the cable diameter. After comparing with the standard cable diameter, analyze whether the cable is normal or there is a defect in the joint or the insulation layer. In this paper, the edge detection algorithm is used to realize the dynamic detection of thermocouple cable defects. Improve the reliability of the products, improve the mechanization of the production process, and in a result of labor saving.

This paper presents a high spatial resolution remote sensing image segmentation method by combining quadtree with minimum spanning tree. Firstly, the improved quadtree segmentation algorithm is used to divide the image iteratively into many over-segmented objects, which greatly facilitates the selection of initial segmentation parameters. Then the improved Morton coding is used to construct the spatial index of the generated over-segmented object and form the region adjacency relation. Combine spectral and texture features, the similarity between adjacent regions is calculated and the region merging criterion is constructed. Based on the idea of minimum spanning tree, the over-segmented objects are merged to generate multiple minimum spanning trees. During that process, the number of minimum spanning trees can be controlled to obtain ideal segmentation results. Compared with two other segmentation algorithms, the method proposed in this paper is more convenient to select segmentation parameters and has certain improvement in segmentation accuracy and object integrity of segmentation results.

Due to the uncertainties of manual operation, it is difficult to ensure that each joint is available for the welded joint of the thermocouple wire. The size of the thermocouple connector is also small, and it is difficult to distinguish the defect of the connector only by the human eye. This paper proposes a method of computer vision, using the camera to automatically identify defects in the thermocouple connector. Use the Candy edge detection algorithm to find the region of interest, and calculate the parameters such as the diameter of the welding joint through functions such as HoughCircles transform algorithm. Finally, the defects of the welding joint are determined according to such parameters. In this paper, the thermocouple welding joint defect detection algorithm is used to realize the automatic detection of thermocouple welding joint defect. It improves the quality of thermocouple welding joint, and the efficiency of the test personnel is also improved.

In the optical interferometry fields, the phase extracted by the arctangent function is a 2π-wrapped phase, it is necessary to carry out the phase unwrapping to obtain a correct continuous phase distribution. However the undersampled phase occurs due to too low sampling frequency and higher fringe density, thus the common unwrapping algorithms will fail. Aimed at the undersampled problem, a phase unwrapping algorithm based on lateral shearing and zooming is presented in the paper. The algorithm combines least square phase unwrapping based on second lateral shearing and bicubic interpolation to obtain larger anti-undersampled range. Taking peaks function as object, the anti-undersampled ranges are analyzed for different phase unwrapping algorithms. It can be shown that the presented algorithm retrieves the continuous phase of 2750 times the peaks function. The algorithm can provide a phase unwrapping solution for the serious undersampled phase, and the analyses of anti-undersampled capability for different phase unwrapping algorithms also are as a reference for future measurement.

Raman spectroscopy provides information about the structure, functional groups and environment of the molecules in the samples, and is widely used in various application areas including chemical analysis, biological processes, environmental and food sciences etc., because of its features of rapidness and non-destruction. The processing and analysis of Raman spectrum is required to extract useful information from original spectrum. For each individual spectrum, a multitude of preprocessing algorithms are required to eliminate effects of unwanted signals such as fluorescence, Mie scattering, detector noise, calibration errors, cosmic rays, laser power fluctuations, and other distortions. Among common methods, Moving Window Average, Moving Window Median and Savitzky-Golay (SG) filter require to set the length of the window, Wavelet based method requires to choose the appropriate Wavelet family, thresholds, and scales, thus the methods mentioned above is not applicable for fully automated data processing and qualitative analysis of handheld Raman spectroscopy. This paper proposes a multi-scale wavelet thresholding denoising algorithm (MWTD). The Raman signal is decomposed into different scales (multi resolution), each scale (resolution) gives different frequency-related information contained in the Raman signal. As noise (high frequency) related frequencies are different compared with genuine Raman bands (mid frequency), at an optimum resolution appropriate thresholds can be applied to eliminate noise. After thresholding (removing) the noise, the corrected Raman signal can be obtained by the Inverse Wavelet Transform. Both simulated and experimental data are used to evaluate the performance of the MWTD algorithm. The results demonstrate that the proposed MWTD method is superior to the hard/soft threshold and Savitzky-Golay (SG) methods in improving SNR, and can effectively eliminate the spectral noise and retain important detail features in the signal. When processing large datasets, a fully automated algorithm such as MWTD would be desirable as it is not required to set any parameters. Thus, the proposed MWTD method is more suitable for the preprocessing before the spectral data modeling and has a better application in the spectroscopic analysis.

The atmosphere calibrated airborne and space borne hyperspectral images are the HDRF of canopy. The spatial nonuniformity of HDRF may result in inversion errors of the heavy metal stressing. In this paper, the HDRF of copper stressed plant samples under different illumination conditions was acquired with the laboratory hyperspectral simulation system called MHRS2F. The difference between the HDRF of canopy and the BCRF of leaves was firstly discussed. Then the changes of spatial distribution of the HDRF for different copper concentrations and illumination conditions were discussed. At last, the sensitivity of various vegetation indices to illumination and observation directions was compared. By comparing the prediction accuracy of different vegetation indices on different observation directions and illumination conditions, the HVI and mRENDVI were found to be more stable and accurate.

Imaging through scattering layers plays an important role in the field of optical imaging. Because of its characteristics, we can observe some targets that are invisible or unobservable. Now, it is a simple and effective way to process images of scattering layers by autocorrelation. However, due to the memory effect and the limitation of the acquisition environment, imaging through scattering layers still lacks the ability to accurately detect unknown objects. In this paper, we analyzed the influence of memory effects and actual acquisition environment on speckle correlations imaging. By controlling the various variables of the experimental device and the image processing, different experimental images and restoration results of the images are obtained. The memory effects control the optical thickness of the scattering layer, the size of the target, and the distance from the target to the scattering layer. There must be appropriate experimental parameter settings to meet the memory effect requirements. In addition, the selection of the position of the image acquisition device determines the degree of dispersion of the speckle. Image processing is mainly for the filtering of space domain and frequency domain, and for changes in constraints in Hybrid Input-Output algorithms. Finally, comparing the influence of all the parameters on the final restored image, the reasonable acquisition scheme and image processing scheme for different targets and scattering media can be obtained. It has reference and guiding significance for the application of imaging through scattering layers via speckle correlations.

It is very difficult to directly measure the effective emissivity of a blackbody cavity, which is generally calculated by the indirect methods. Different calculation methods have certain differences. In this paper, the STEEP software, TRACEPRO software and approximation formula are used to calculate the effective emissivity of the blackbody cavity for the three commonly used blackbody models of integrating sphere, cylindrical and cylindrical-conical, and Compare the results of emissivity greater than 0.99. The results show that the consistency of the results of the three methods in calculating the cylindrical type and the integrating sphere type blackbody can be better than 0.03%. When calculating the conical black body, the difference between the results of the approximate formula and the other two schemes is less than 0.07%. The calculation results of the scheme tend to be consistent with the increase of the emissivity of the model.

In order to improve the accuracy of polarization target detection, the multi-parameter polarization contrast model is proposed after analyzing the typical polarization features of the polarization images. It utilizes both of the polarization degree and the polarization angle parameters. Then the fast polarizer angle detection method is designed according to this model to calculate and drive the motor to rotate the polarizer to the most appropriate deviation angle so as to maximize the contrast between the target and the background. Experimental results show that the proposed method can improve the contrast between the target and the background in the polarized image significantly, which makes the polarization detection more efficiently and lays a foundation for detecting the moving targets.

In this paper, we introduce the working principle of dispersion imaging spectrometer, describe the system composition of imaging spectrometer. Through analysis and comparison, we select the Cerny-Turner for the optical path structure of spectrometer. Then, we simulate and optimize the initial structure of the optical system of imaging spectrometer by using optical design software. At the same time, we use the cylinder reflector with low cost and easy processing and detection to correct astigmatism. Finally, we designe an optical system of vis-shorter-wave infrared imaging spectrometer with a spectral range of 400~1000nm. The contrast of full-band and full-field MTF of imaging spectrometer at 25lp is better than 0.3. After these, we design the mechanical structure of the optical system of the optimized imaging spectrometer by using the mechanical structure design software. Then, we use 3D printer to print various structural parts to develope a high resolution imaging spectrometer. Finally, we carried out an experiment by the kunyu river in haidian district, Beijing, which proved that the imaging spectrum technology can be used in the monitoring and research of water pollution and vegetation in rivers and wetlands.

In coherent test system like sensing, imaging and communication, the coherence of the light source is very important. Our recent advances focusing on the measurement of noise characterization of highly coherent laser and its applications in the coherent test system are reviewed. Using the noise spectrum and statistic characterization to represent the coherence is proposed. And the parallel 3-step phase-shift measurement method from 500nm to 2000nm based on the 120-degree phase difference fiber interferometer is also proposed to measure the noise characterization. Some kinds of typical highly coherent laser applied in the coherent sensing and imaging system are measured. And the impact of their performance on the application system are also discussed.

In order to meet the demands of fabric defect detection under different lighting conditions, the multi-scale Retinex algorithm is proposed as preprocessing algorithm to limit the influence of lighting change on subsequent processing to a certain degree. Firstly, the fabric defect simulation database under complex lighting conditions is produced by rotating, flipping and transforming the data based on traditional TILDA fabric texture database. Aiming at the phenomenon of the obvious brightness changes between different images in the database and the more complicated illumination environment, the multi-scale Retinex algorithm as the preprocessing is used by logarithmically transforming the given input image and estimating the incident image in this paper. The input image and the estimated incident image are reflected images, which limits the influence of illumination changes on subsequent processing to a certain extent. The comparative experiments show that dynamic range compression, color constancy, edge enhancement and a balance between the three aspects can be achieved by the multi-scale Retinex algorithm at the same time. The experimental results show that the multi-scale Retinex algorithm is robust, and the local details of the processed image will be well maintained. The image information entropy and contrast is increased by 30%, and average gradient is increased by nearly 40%. Simultaneously, the change of light and noise will be limited to a certain degree, and high-quality fabric image under different illumination conditions can be obtained effectively.

In order to accurately monitor the working temperature of turbine blades, a signal processing technology for simulated blades is proposed in the paper. The method includes four steps: obtain the function relationship of voltage and temperature by fitting; segment the waveform of each circle by synchronization signals; filter the noise signal by Fast Fourier Transformation (FFT) and Butterworth low-pass filter; then restructure two-dimensional temperature by mapping temperature data into the polar coordinates. Finally, compared with the true temperature measured by thermocouple, the absolute temperature error after signal processing is no more than 4oC at temperatures range from 550oC to 1000oC, providing valuable guidelines for condition monitoring and fault diagnosis.

Endmember extraction technology of ballistic missile is an important research content in spectral remote sensing, which can effectively solve the mix pixel problem. This paper starts with the background requirement of near space and research content of mixed spectral endmember extraction. The algorithm of spectral endmember extraction based on non-negative matrix factorization for ballistic missile in near space background is proposed and analyzed. The simulation results show that the proposed algorithm demonstrates the good performance in the condition of random mixing mode and correlation mixing mode.

To solve the problem of blurred edges in some color images, a color image enhancement algorithm based on edge extraction is proposed. This algorithm is carried out in YUV space. The improved Laplace operator is used to extract image edge, which improves the clarity of edge while preserving image color. In addition, this algorithm can also improve the edge clarity of blurred and noisy images.

In optical three-dimensional imaging and measurement based on fringe projection, wrapped phases may cause a problem of ambiguity, which can be overcome by phase unwrapping generally. In this paper, a method for absolute phase unwrapping using light-field imaging is reported. In a recorded light field under structured illumination, i.e., a structured light field, a wrapped phase-encoded field was retrieved and resampled in diverse image planes associated with several possible fringe orders in a measurement volume. By leveraging phase consistency constraint, the resampled wrapped phase-encoded field correct fringe orders could be determined to unwrap phase without any additional encoding information.

The study of imaging through scattering media especially 3D imaging is of great significance in many fields such as biomedical imaging. Recently, deep learning has been widely used in the field of information processing with its remarkable performance. In this paper, we proposed a method of three - dimensional imaging through scattering media based on deep learning. This method uses the deep neural network to process the information captured by the light field imaging system based on the microlens array, recovering the no-scattering 4D light field information, and then realize three-dimensional reconstruction by using the processed light field information. Deep learning method requires a large number of samples. But in many environments, it is difficult to obtain a large number of three-dimensional samples through experiment. To solve this crucial problem, we use incoherent light propagation model to simulate the light field propagation and generate samples which contains three-dimensional information through simulation. In this paper, we simulated the propagation of radiation emitted from objects behind a single layer of weak scattering media, generated a large number of samples of 4D light field information by simulation, trained the neural network and processed the test data set generated by simulation, and we realized the deblurring of the light field information which contains information of multiple layers of flat semitransparent objects, which could be used to realize the 3D reconstruction.

Astronomical observation and spatial target surveillance applications often require mosaic processing of starry images acquired by multiple image sensors to expand the Fields of View (FOV) or improve the resolutions. Due to the low SNR (Signal-to-Noise Ratio), lack of star point texture information and vulnerability of atmospheric turbulence of the starry image properties, traditional mosaic methods are prone to failures during feature point extraction. In this paper, Spatio-Temporal Context (STC) filtering is introduced as the preprocessing procedure to suppress the background interferences. We have improved the classical STC filtering and expands it into multi-scale space combining with Rolling-Guidance Filtering Algorithm (RGFA). Making full use of the fine edge-preserving feature of RGFA, the time-variant or spatial variant interference and noise in the background, such as glimmer stars, night clouds, sensor response noise, etc, are suppressed while the profiles of the target star points are enhanced and easy to extract their centroids. Then, we produced the feature description of the star-point sets via threshold segmentation and morphological algorithms based on geometric invariant cost function for the input image pairs to be stitched. After Random Sample Consensus (RANSAC) processing, the mismatched feature point pairs in the star-point sets are excluded. The subsequent procedures of the registration parameter calculation, image fusion and parallax correction processing are adopted to complete the mosaic processing. The results of digital simulation and practical processing show that the proposed method for the multichannel sequence starry images with the low SNR and complex backgrounds can extract feature points more precisely and more robustly comparing with the traditional methods. So, it is suitable for the large FOV spatial observation or surveillance applications.

Traditional approaches for remote sensing image segmentation are mature in certain respect, as the new spaceborne technology of continuous observation satellite video emerges, it arises a new demand for moving object detection from this new data source. In the field of computer vision, deep learning technique has achieved outstanding performance for general images. In the research, a deep learning based method is introduced and several modifications are made in the processing steps. Faster Convolutional Neural Network (Faster CNN) algorithm is selected as the basal pipeline and conditional random field is used to generate finer detail proposals. After enforced iterations, the computed extraction result of ships in remote sensing satellite video data is compared with original Faster CNN method which demonstrates an improved target detection output in different tests.

The dim target tracking is essential for the spatial surveillance system. Considering that the starry image sequences acquired by imaging sensors often has low Signal-to-Noise Ratio (SNR), the brightness of a spatial target is often susceptible to the background interferences, such as the night clouds and the atmospheric turbulence, etc, and become dim and instable, its shape and profile is also blurred and lack of texture information. In order to extract the target from background, Spatio-Temporal Context Model (STCM) based filtering theory is applied in this paper and used to improve the traditional Kernelized-Correlation-Filter (KCF) target tracking method. It introduces a spatial weighting function that can pre-enhance the point target and suppresses the background interferences. So the tracking drift phenomenon is relieved when the moving object being obstructed temporarily. Considering that L1 regularization is easier to obtain sparse solutions and L2 regularization has smoothness property, the regularization function of the regressive classifiers in KCF target tracking method is renewed by using variable L1 or L2 regularization instead. The index of regularization in the improved regression model is a piecewise function, which is determined by the cost function during learning period that can distinguish the target star point from the background point by using the characteristics of points (such as brightness, etc.)The numeral simulation and actual processing results show that, comparing with the traditional Kernelized- Correlation-Filter (KCF) methods, the proposed method owns more robustness and precision in the starry images with low signal-to-noise ratio and complex background.

In the spatial target surveillance and astronomical observation applications, image matching processing is the key procedure for the multi-temporal starry images or the multi-channel starry images acquired by different imaging sensors. However, the starry images obtained often have low Signal-to-Noise Ratios (SNR), the light intensities of the target stars or spacecrafts in them are vulnerable to background interferences, such as the atmospheric turbulence and the night clouds, etc., and become dim and instable. With the weak texture information of the target stars, all the influences make the feature point extraction quite difficult. In this paper, a new type of image matching method based on the description of Multi-scale Geometric Invariant Features (MGIF) is proposed, which uses the Rolling Guidance Filter (RGF) to perform preprocessing for the input images. By virtue of the excellent edge-preserving performance of the Joint Bilateral Filter in RGF, the integrities of contour profile of the star points are guaranteed effectively while the interference and other noise in the background are suppressed. Then the segmented and morphology methods are applied to extract star points and get the centroid of star points to form the feature point constellation. Considering the cross ratio of two lines in projection transformation model of image matching is a geometric invariant, a multi-scale geometric invariants based function, which uses the scaling of RGF as a reference to describe the relative spatial positions of matching points more accurately, is constructed to evaluate the level of similarity between star points according to the relative position of each points in the constellation. Subsequently, Random Sample Consensus（RANSAC）method is adopted to remove the mismatching star points and calculate the rigid transform matrix and other registration parameters. Digital simulation and practical processing results demonstrate that the proposed method can achieve higher matching accuracy and robustness for the starry images with low SNR and complex backgrounds.

Conventional optical microscopy provides only intensity images, for which the contrast is induced by fluorescence or the absorption of the sample on the illumination light. Yet, the phase, polarization, and spectrum information of the sample is lost. Meanwhile, limited by design, conventional optical microscopy suffers from the conflict between spatial resolution and field of view (FOV). Modulated illuminations based computational microscopy (CM), which joints front-end optics and post-detection signal processing can, in general, extend the capability of conventional microscopy; for example, it allows the acquisition of the intensity, phase, polarization information, and enhance the spatial resolution within a large FOV. In this paper, modulated illumination based CM was exploited for implementation of phase imaging, resolution enhancement, dual-modality imaging. First, modulated illumination based CM provides quantitative amplitude and phase images, revealing the 3D shape and the inner structure of transparent or translucent samples in the absence of fluorescent labeling. Second, pupil-segmentation based CM measures the aberration of focus modulation microscopy (FMM). Hence, the resolution and SNR of FMM was enhanced after the aberration compensation. Third, phase and fluorescence dualmodality imaging was implemented in confocal laser scanning microscopy (CLSM) by extending the depth of field (DOF) of the CLSM system with a tunable acoustic gradient index of refraction (TAG) lens, providing complementary information (structural/functional) with pixel-to-pixel correspondence for the same sample. Furthermore, the combination of the two imaging modalities enables standalone determination of the refractive index of live cells.

As the most important greenhouse gas, CO₂ has caused a lot of climate change problems. Therefore, the measurement of CO₂ is of great significance. Based on the tunable diode laser absorption spectroscopy (TDLAS), the effect of temperature on the absorption spectrum of CO₂ was studied. A DFB laser with the center wavelength of 1430 nm was selected. The absorption line of CO₂ near 1432 nm at different temperatures (293-373 K, interval 10K) was measured by wavelength modulation spectroscopy (WMS). By comparing S-G filtering method, FFT filtering method and P-F filtering method, the spectral data was preprocessed, and the optimal preprocessing method was determined. The measurement error caused by noise was reduced effectively, the second harmonic signal of CO₂ at different temperatures was obtained, and the influence of temperature on the absorption spectrum of CO₂ was analyzed. The goodness of fit of temperature and signal strength under different orders was evaluated. Finally, the second-order polynomial fitting was chosen as the optimal model of CO₂. The experimental results complement and improve the existing database to ensure the accuracy of CO₂ inversion, which has an important reference significance for the measurement of CO₂ in practical applications.

In general, the image processing part is placed on the personal computer (PC) for processing. However, with the amount of data to be processed increases dramatically, the requirement for data processing speed is increasing. Further, a real-time imaging system using Field-Programmable Gate Array (FPGA) is implemented, which is designed for a large-array CMOS camera based on USB3.0 interface. Given the advantages of FPGA parallel processing, this design will implement sensor image processing pipeline (IPP) with the FPGA. At the same time, the high speed storage device double data rate (DDR3) is used in the system to cache image data. A Xilinx FPGA is the core processing unit of the entire system, and it completes all the functional modules, including the CMOS driving, data transmission, real-time image stitching and image caching. FPGA-based IPP and DDR3 frame buffer application can improve the video frame rate. Finally, the imaging experimental results show that the large-array CMOS real-time imaging system has a display resolution of 7920*5432 with a stable frame rate of 8.5 fps, and achieves a good balance between speed and efficiency.

In order to select more suitable characteristic lines as the analysis lines of the elements, laser-induced breakdown spectroscopy is used to detect the self-made soil sample doped with Cd and Fe elements. Using Savitzky-Golay convolution smoothing method to preprocessed spectral data. Select the common line Cd I: 288.122nm, 346.62 nm, Fe I: 357.001 nm, 363.146 nm to establish the calibration curve by external standard method, calculate the detection limit ， the influences of atomic configuration and angular momentum on the calibration curve and detection limit of the spectral transition level are studied. The study shows that the detection limit calculated by the analysis lines with high SNR are relatively low and more sensitive to the detection of elements; The SNR of the analysis lines with the same atomic configuration of the transition show the same trend with the change of element content. The analytical lines Cd I: 346.62 nm (2-1) and Fe I: 363.146 nm (4-3) with relatively large angular momentum are better excitation, and the correlation coefficient R2 of the obtained calibration curves are relatively higher, the limit of detection is smaller than the Cd I: 288.122 nm (1-1) and Fe I: 357.001 nm (1-1), the lines with large angular momentum are more suitable for element detection.

The positioning accuracy of the PDR based on the smartphone is relatively low due to the accumulative error caused by the heading in inertial navigation. In order to resolve this problem, in this paper, we use the solution that fusing the heading which is measured by gyroscope and orientation sensor. In addition, we propose a new fusion method which is realized by the radial basis function neural network and compare the fusion positioning results with the Kalman filter and Back Propagation neural network. The experimental results shows that the positioning error corresponding to 80% confidence interval processed by the radial basis function neural network is only 8.18cm, while the results of Kalman filter and Back Propagation neural network are 34 cm and 22.54 cm, respectively. The experimental results show that the proposed method has the higher positioning accuracy than the traditional Kalman filter method and Back Propagation neural network. These experimental results demonstrate that the radial basis function neural network can be used in the indoor high-precision PDR.