The traditional moving object detection algorithms cannot work when both background and objects are in moving state. A method of infrared image moving object detection is proposed on moving background. First a selecting method of image characteristic block based on frame differences is proposed and the improved mutual information measure is combined to match background. Then the matched images are dealt with frame differences and the frame differences image is segmented by the threshold which is gotten based on image characteristic block. The background noise is wiped off by using morphology method. At last, the object detection is completed by image projection. Experiments show that the proposed algorithm can eliminate background motion effectively and detect the moving objects on the moving background accurately.

The problems of conventional Maximum Likelihood Estimation (MLE) algorithm are addressed for passive localization and an improved MLE passive localization algorithm is presented. Firstly, we estimate an initial target position using least square method. Moreover, in order to adapt the measurement error, the square root of the difference between the estimated position and the sensor’s position is used as the approximate covariance matrix for measurement error. Then, a weighted least square formula is employed to estimate a new position. Finally, we regard the estimation value as the necessary initial value, and employ the conventional MLE to calculate the final results. The improved algorithm have some advantages, i.e., it does not need to set an initial target position, its localization results do not diverge easily, its computational complexity is low, and it has the same level in accuracy as that of conventional MLE. Experimental results show that the proposed algorithm is effective.

To satisfy time-sensitivity of inshore infrared target segmentation, we improved SBGFRLS segmentation model by analyzing the features of inshore infrared image. First, visual attention model based on phase spectrum is used to replace the locating method of original contour of SBGFRLS model which decreased segmentation field and iterations. Second, the peak values of grey level histogram of inshore infrared image are used to replace the parameters of SBGFRLS model which saved calculation time. Our experiments prove that the improved algorithm is feasible and greatly improves the speed of segmentation compared with SBGFRLS model and other common segmentation method.

As to the human body tracking problem under the weak light environment, human body tracking algorithm of ant colony detection with pigment information is presented. The character of ant colony optimization algorithm was used to detect the monitoring initial frame image object edge, and in the initial frame image to gather on the edge of the ant colony body, pigment information is marked. By using pigments ant colony detected with pheromone ant colony and toward pheromone ant colony gather, the purpose of tracking for human movement at night is achieved. The experimental data analysis and simulation results show that the human tracking algorithms for human movement under the low light environment is effective, and also has certain robustness for an image with noise.

By analyzing the principle of polarization detection, a method that was applied to recognize camouflage targets with their polarization information was presented. A multi-spectral polarization system was designed and the processing software was programmed. Then the experiment about the iron plate and camouflage plates placed on typical woodland background were conducted by the system at 400～1 000 nm. The results of image processing and analysis indicate that the polarization characteristics of camouflage plates and background are different. By comparing with natural background, the polarization characteristics of camouflage plates are obvious. The result shows that the classical camouflage target can be detected and identified effectively by the spectral polarization technique. Consequently, compared with the traditional intensity detection, the spectral polarization detection technique is of significance to improve the veracity of detection of camouflage targets in mixed background.

A novel dense matching algorithm based on disparity growing and trifocal tensor is proposed because of the problem of dense matching. SIFT algorithm is used for the feature extraction of three-view. Then, the coarse matching is obtained by using the method of Euclidean distance. For removing the part incorrect matching of coarse matching set, gradient principal direction angle difference histogram of key point is used. Trifocal tensor and initial value of disparity which is used for directing dense matching is calculated by three-view matching. Root points which are selected from two-view matching points are used for disparity growing. Three-view dense matching is obtained by using point correspondence of trifocal tensor the same time. Then, initial value of disparity is used for improving the precision of the extraction. The experimental results indicate that the method get the accurate and dense matching in widely baseline case. Dense disparity map can be obtained.

The cannon chamber detection is an important task in the gun bore maintenance process, and it has direct relationship with the firing accuracy and security. The features of cannon chamber images obtained from grin lens array are analyzed, and an algorithm based on the phase-correlation and log polar is presented, which can realize mosaic image fast and seamlessly. Phase correlation algorithm can only match the images with shift invariability, and would face difficulty when dealing with motion model，where either scaling or rotation of the matched images is presented. Log-polarimage transform can convert the scale and rotation changes in Descartes coordinate to shift changes in log-polar coordinate. This method calculates the normalized peak of phase correlation of cross-power spectrum, and gets the rotation and scale between the two conjoint images using the peak coordination. Images are matched based on these parameters, and then chamber panoramic images are obtained by using method of pixels maximum value. Experimental results show that the proposed method can get seamless, clear and full image of gun bore.

In view of two-dimensional distribution reconstruction of the atmospheric trace gas concentration, a new method, passive Differential Optical Absorption Spectroscopy (DOAS) tomography for atmospheric trace gas distribution reconstruction, is put forward. Two MAX–DOAS is applied to scan SO2 plume, the DOAS technique is used to inverse gas column concentration, and algebra iterative algorithm (ART) is applied to extract information of the concentration distribution. Moreover, the Gaussian plume is adopted as reconstruction object, and the method of numerical simulation is adopted to study and analyze reconstruction effect. The analysis shows that the reconstruction result is consistent basically with the model, so the method of gas distribution measurement is feasible. In order to study influence of error in the actual measurement on reconstruction results, the different random error is added to projection data. The study shows that a reconstruction result is insensitive to random error in projection data.

A novel three-dimensional shape measurement method is presented for low current tungsten inert gas welding pool surface, based on fringe reflection and Fourier Transform Profilometry (FTP). The relationship between surface height and phase variation has been demonstrated. Reflection character of liquid metal on welding pool surface and scattering function result from molten pool concussion have been discussed with the purpose of taking the picture of reflected deformed laser stripes by means of pool height measurement device. The obtained deformed laser stripes of low current tungsten inert gas weld pool surface were processed by computer to acquire pool surface height. An experiment has been done to demonstrate that Fourier transform profilometry is suitable for three-dimensional shape measurement of three-dimensional shape measurement.

As to the low dispersion glass H-FK61, the dispersion value is small, and its variation range is narrow, which have a very strict requirement to the measuring method. Under such circumstances, the formulae of three minimum angles of deviation are derived, and its accuracy is analyzed in comparison with traditional minimum angle of deviation method. The analysis shows that under the same condition, the accuracy of three minimum angle of deviation is higher than that of the minimum angle of deviation. Finally, we verify it at the goniometer, and solve the measuring problem of H-FK61, which lay a foundation for the measuring of another special glass FCD100.

Former method to detect the wavefront of diffraction grating was based on special system .It is hard to avoid errors caused during setting up optical system because so many devices are used. In this work, the measurement system based on ZYGO GPI. XP/DX interferometer detect the wavefront of plane grating and concave grating, which expanded the applications of ZYGO interferometer. The principle of the system is traditional interference method. ZYGO interferometer is used in concert with standard lens and mirrors as reference device. The precision of reference devices approach λ/40 and the precision of interfere cavity is λ/20. The ±1 order diffraction wavefront and its interference fringe of plane grating are gotten. Two kinds of testing concave grating wavefront were compared. Analysis shows that the new design is simpler and easier to set up compared to past devices.

Various backlight algorithms are analyzed comprehensively in the article. Moreover, considering the real-time characteristics of LCD (liquid crystal display), a new algorithm based on dynamic threshold is proposed. The algorithm is based on the average backlight, combines with the proportion of the image details to adjust the backlight, and can maintain the quality of the image and reduce the power consumption effectively.

In order to extract the road traffic flow information using UAV images, it is necessary to calculate the location of the road area and generate the road template for road tracking. Real-time calculation requires automated seed points extraction for identifying road and generating tracing template. A one-dimensional unmarked watershed segmentation method is proposed, which needs to calculate using the simple image feature, and don't need manual seed points mark. Firstly, by recognizing sectional extremum in the profile of source image road, the valley regions are extracted between extremums and the similarity is calculated through the region. Then, combine the nearly similar regions, get the segmentation results. Secondly, rank the segmentation regions, search radiology in the ranked regions, and identify the road area and calculate road track template. Experimental results show that the method need less manual intervention and more efficiency than the Canny edge extraction and Hough line extraction under the same conditions. By fitting the road edge points by the least squares method to get the direction of the road, the calculation of road direction is accurate, and the calculated tracking template is consistent with the actual that.

Image segmentation is a hot point in the research field of automatic target recognition of SAR image. In order to segment the MSTAR image automatically, a new adaptive method is proposed. Firstly, 2D Gabor filters with various orientations and scales are used to enhance the original image, which can effectually reduce speckle noise in the background, and smooth the interior of the homogeneous regions. Then the analysis of the statistical characteristics of the enhanced image is made, based on which the adaptive thresholding rules is presented for the automatically segmentation of the images. Experiment results with the MSTAR images indicate that the algorithm presented has advantage of segmentation accuracy, calculation efficiency and noise robustness over the traditional methods, such as OTSU, FCM and MRF.

In the processing of Super-resolution (SR) based on sparse representation, when a testing image is sparsely coded, the coding residual doesn’t simply follow Gaussian or Laplacian distribution. In the maximum likelihood estimation principle, a mixture Gaussian sparse coding model is proposed to solve this problem. Firstly, a weighted l2 norm function is defined to approximate the optimization problem, different weight is defined for different coding residual, and the function is solved by iteratively reweighed sparse coding algorithm. Then the SR model and algorithm are established based on the proposed model, the isomorphic high/low resolution over-complete dictionaries are trained and the sparse representation coefficients of the testing image is learned by the proposed method. At last, the reconstruct method is mended to improve the robustness to noise. The experimental results demonstrate the effectiveness of the proposed method.

The current method of SPECT/CT color medical image registration uses the location marker in vitro generally, such as SPECT/CT color medical image, but using the external-marker stent fixation with patient for examination is complicated relatively. A method of calculating registration parameters such as rotation angle and translation amount is proposed based on Quaternion color model. Firstly, the color medical image is represented as Quaternion vector matrix form. On the basis of this, a calculation method of Quaternion geometric moment is put forward about SPECT/CT color medical image. According to the distribution of Quaternion image quality, the relative rotation angle of two-mode image is obtained by using rotational inertia properties of Quaternion image. Then, the translation amount between two images is determined to realize registration by using Euclidean distance measure of mass center. The SPECT/CT color medical images are used in registration experiment. Good registration results are obtained, and this method is proved to be of effectiveness and feasibility.

Human action recognition in the video sequence have become a hot research topic in computer vision field. In order to extract the contour feature of the human’s behavior sequence more effectively, a new algorithm for human action recognition based on Local Binary Pattern (LBP) is proposed. Firstly, background subtraction algorithm is used to extract the complete human motion sequence in the video, and the LBP operators are used to calculate the samples’ LBP feature space which is composed of the motion sequences’ LBP feature spectrum. Then, the behavior feature is generated by k-means clustering method. Finally, the Hidden Markov Model (HMM) is adopted for the classification. During the experiment, the test experiment is performed in the two public behavior databases respectively, and the average recognition rate can reach more than 85%. The intersection of the two databases experimental results shows that the proposed algorithm has certain robustness.

Based on the principal component analysis, 1ocality preserving projection and kernel mapping theory, kernel within-class neighborhood preserving orthogonal algorithm with global structure is proposed. The object function ofproposed method takes the global and local information of the samples into account. At the same time, by using nonlinear mapping and the orthogonal restriction of base vectors, this algorithm can extract more effective classification features. Being unknown nonlinear function, the object function can not be directly calculated. But according to the kernel mapping theory, the equivalent object function of the proposed algorithm is established in the kernel feature space. The deducing process and solving steps are given. Experimental results on face database demonstrate the effectiveness of the proposed method.

A part-to-part curve matching approach for occluded object recognition is proposed since the global curve matching methods perform infeasible in occluded object recognition. First, a local curve feature descriptor called Affine Invariant Curve Descriptor (AICD) is constructed to represent the local shape of a curve. It is affine invariant based on curve affine invariant signature and its unsigned integral. Then, combining AICD with a curve segmentation strategy based on curvature zero crossing point, an affine invariant curve matching algorithm is developed. Experiments demonstrate that the proposed algorithm exhibits effectiveness in object recognition with affine deformation and partial occlusion and outperforms the state-of-art curve matching methods.

Nowadays many enterprises lack of strict optical design and optimization to integrated light source, and the mould is opened just based on experience, which has low efficiency and high cost. Non-sequential Monte Carlo ray tracing method and optical simulation software Tracepro are used to do first optical design and optimization of integrated light source. Based on the structure of LED integrated package, a 7 W COB optical model is developed. Then, four variables of chip distances, bottom radius, depths of reflector cup and refractive indexes of silica gels are changed one by one to obtain the most excellent value of each parameter. The experimental results indicate some laws about the impact of chip distances, bottom radius, depths of reflector cup and refractive indexes of silica gels on luminous efficiency and light intensity distribution, which has guiding significance on the actual production.

According to the fact that it is difficult to satisfy the requirement of high structure stiffness and thermal stability with traditional support methods for large aperture rectangular reflector under self-gravity and large temperature changes, a spherical hinge support structure is introduced to effectively improve the surface accuracy of large aperture rectangular reflector under self-gravity and thermal load. Theoretical model analysis was conducted and analytic solutions were derived which could be useful to guide the optimization design of the support structure and predict the stiffness at the design phase. Finite Element Analysis is conducted for the spherical hinge bipod support of a 540×437 reflector. The result indicates that this kind of support structure can effectively improve the assembles’ thermal stability and stiffness and well satisfy the requirement.

Some instruments such as motion stages and shutters in optical holographic experiments are commonly used with automatic programming control. The automatic control is commonly achieved based on instruments controlling system with PC as major component，which is more complex, high costs and less compact. In this paper, instruments controlling system for optical holographic experiments is designed and implemented based on ARM Embedded System. The system communicates with instruments via standardized interfaces and interacts with users by using visual interfaces. The whole system is integrated by printed-circuit boards. The instrument controlling system can implement auto control to shutters, motion stages, and power meters. Visual interfaces for control and display are designed and achieved.

We propose a novel metamaterial structure process according to the transmission property of the Silicon-on-insulator (SOI) wafer. We also suggest an appropriate structure for the Micro-electro-mechanical System (MEMS) technique to achieve larger frequency modulation. This structure was assembled by a Split Ring Resonator (SRR) with a metal mesh within a unit cell. Our experimental studies on the composite structure indicate that the coupling of the metal mesh and the SRR significantly contribute to transparency at the terahertz range. Moreover, we experimentally demonstrated the performance of this structure by changing the relative positions of the SRR and the metal mesh. The unique resonance characteristics of this composite structure make it an excellent candidate for developing tunable terahertz components via integration with the MEMS technology