An adaptive method to determine chip rate is proposed for laser communications based on diffuse in a close space. Compared with conventional laser communications in free space, the method adds a test process of chip rate in the beginning phase of communication. In the process, equipment 1 and equipment 2 communicate using fixed low chip rate, and a series of test data composed of alternant appearance of signal “1” and signal “0” is transmitted from equipment 1 to equipment 2. In virtue of the accepted data in equipment 2, the persistence time of signal “1” and appropriate chip rate of communication in existing circumstances can be calculated. After equipment 2 transmits the chip rate to equipment 1, the two equipments will communicate using the new chip rate. There are two merits about the laser diffuse communication system using the aforesaid chip rate. The first is neither resulting in errors due to delay of laser diffuse nor wasting communication resource due to low chip rate, and the second is that communication circumstance needs not to be considered nor the chip rate can be obtained automatically by software.

According to the definition of satellite’s magnitude, for geosynchronous orbit space satellite, the theoretical computing model of star magnitude was built. Space vectors among sun, observation station and target satellite were derived to calculate the time-space relationship parameters. The simulation curves of magnitude vs. time were drawn based on the model and corresponding parameters. The simulation results were verified to match the experiments ones by using satellite targeting system and astronomical telescope. At last, one geosynchronous satellite with 13.4 magnitudes was observed.

The technology of CCD as a star sensor starlight detector has been very mature, but the bottleneck is becoming exposed. On the basis of parameters of IBIS5-B-1300A, the article analyzes CMOS APS noise effects on position accuracy of star sensor, obtains the maximum star position accuracy which a CMOS APS sensor can achieve and verifies the theoretical analysis results through modeling and simulation. Then two conclusions are drawn. First, when gauss dispersion radius is 0.7 and the windows size is 5×5, we can obtain optimized subpixel accuracy; Second, Fixed Pattern Noise (FPN) has the greatest effect on the accuracy of centroid algorithm. This article provides a theoretical basis and support for the selection of the starlight detector for a star sensor system.

A dynamic dictionary learning method was proposed for automatic target recognition in Synthetic ApertureRadar (SAR) images. The new method decreased the size of a dictionary by erasing the useless or high-correlation atoms in the dictionary under a deleting cost. It added an atom by decomposing the residue data matrix under an adding cost. The performance of signal representation of a dictionary could be improved in the dynamic learning procedure. Experiments based on MSTAR database show that the proposed method can converge to a proper-size dictionary no matter how large the original size is. The proposed method was demonstrated to have good and robust performance of SAR target recognition in the trials.

It is hard to detect the dim small target with few pixels and low intensity in the image of complex background. According to the intensity features of dim small target, a scheme based on histogram adjustment and hit-miss transform is presented to detect dim small target. In this method, histogram adjustment is adopted to enhance the signal-cluster-ratio, and hit-miss transform is used to get the suspicious targets. The experiment shows that the method is efficient in detecting dim small target with complex background.

To analyze the temperature deformation of photo-electronic theodolite mechanical structure, the entity model of theodolite mechanical structure was established, and stable-state thermal analysis of mechanical structure was done by using the software Ansys/Workbench. Moreover, the temperature distribution of mechanical structure, temperaturedeformation of mechanical structure and the temperature deformation of mechanical structure along Z direction were given. The analysis shows that part of the mechanical structure has a greater deformation under the conditions of thermal inhomogeneity and temperature changes. At 27.938℃, the maximum deformation of mechanical structure is 0.133 mm, the deformation of columns causes horizontal axis inclined 2.8″, and deformation reduces pointing accuracy of theodolite. The analysis provides reference information for structure optimization design, distribution of temperature sensor and local temperature control of photo-electronic theodolite.

A novel two-step phase-shifting profilometry is proposed after the Gram–Schmidt (GS) orthonormalization is introduced into three-dimensional (3D) shape measurement. Firstly, two phase-shifted randomly sinusoidal fringe patterns are projected onto the tested object by Digital-light-processing (DLP) projector. The fringe patterns modulated with theobject’s surface are captured by a CCD camera. Secondly, the background component is eliminated from deformed gratings by averaging method pixel by pixel after the appropriate window size is adopted. Thirdly, an orthonormalized fringe basis is determined from the two fringe patterns using the GS method. Then, the phase is demodulated from the fringe basis. Finally, we have tested the proposed method with real measurement of the human mask. Our algorithm is compared with the other methods. The experimental results are given, which demonstrate the validity of our method.

For the problem of low calculation efficiency of CPU 3D measurement under CCD camera with high resolution, a new fast calculation method is proposed by GPU technology. At the same time, for the four parallel processes of 3D measurement, the method presents realizable specific algorithm processes in GPU. According to the data size and GPU characteristics, a new GPU data structure on device terminal is designed. The experimental results validate that the method effectively utilizes GPU massively parallel processing ability, the measurement time is only 4 seconds in the processing of the five-megapixel data, and the computational efficiency of the method is 160 times of CPU.

To defect detection of Printed Circuit Board (PCB) is an important in the detection domain of computer vision. In order to use reference method to detect photoelectric image of printed-circuit-board, we need to improve accuracy of image registration. The new registration method based on randomized Hough theory and affine transform in the geometric transformations is proposed. The efficiency of finding and finishing image registration has been induced. After the contradistinctive experiments for the detection of photoelectric image of PCB obtained in reality, the detection method proposed in this paper has obvious advantage in the aspects of increasing detection precision, reducing memory space and time.

A precise micro-displacement measurement system based on network communication is introduced. Network communication, FGPA and the linear CCD are adopted in the system. The linear CCD generates 5 000 line data per second. The FPGA processes the data synchronously. The network communication is applied to implement long-distance data transmission and multi-station communication. The hardware system is controlled by the PC control program, which is developed by VB6.0. The whole system demonstrates high reliability, and it is applied in the one-dimensional micro-displacement measurement system with a precision of 100 μm.

A new calibration method of line laser monocular vision sensor based on a new free position and orientation planar target is proposed for the deficiencies in existing methods. Projective invariance of a set of vector product orientations are applied to sort the image control points and then calculate the correspondence between image control points and target control points. Laser control points in each image are extracted by calculating the intersection points of laser stripe and ordered straight lines connecting control points. 3D coordinates of the laser points in camera coordinate frame are calculated by cross ratio invariance principle. Therefore the sensor is precisely calibrated with the optimization calculation of camera parameters and laser plane parameters. Verified by experiments, this calibration method has high-freedom operation, simple calibration process and ideal calibration accuracy.

To achieve accurate measurement of focal length of lens and optical system, the measurement theory based on Talbot-Moiré is analyzed. The system includes a collimated light generator, two Ronchi gratings and an image acquisition system. The theoretical issues and deviation are analyzed. The measurement uncertainty also increases with the increase of the focal length under test. When the focal length of lens under test is 500 m, the measurement uncertainty is 5.23%. The optical system whose focal length is 75 272.80 mm is measured, and its measuring relative deviation is 1.09%.

An autofocus method based on the analog video signal of the microscope CCD (Charge Coupled Device)is presented, which can improve the autofocus range. In many autofocus modes, the high-frequency energy of the video spectrum is taken into account, while ignoring the changes of the low-frequency energy of the video spectrum in the focusing process. In this presentation, the spectrum of the analog video signal was analyzed, including the affection of the different frequency range analog signal on the focusing evaluation curves. Based on the analysis, we obtain the proper frequency range that fit the coarse/fine adjustment of the autofocus process. The resulting system has a wider focusing range while remaining the high accuracy. Experimental results show that the proper filter can extend the autofocus range.Using the 40× microscope objective, the effective autofocus range of the microscope can reach ±500 μm.

Workspace measuring and positioning system is a type of large-size measurement system based on intersection of optical planes. Optical surface was discovered a non-planar face while measuring, which influenced the measuring accuracy of the system. We presented a method of describing optical surface by using ternary quadratic equation. Ternary quadratic equation could be obtained by fitting the discrete points on the optical surface with the method of linear least square. Based on the ternary quadratic equation and orthogonal linearity transformation, we analyzed the shape of the optical surface. By using the method of quadric surface fitting and plane fitting, respectively, with the same group of data, then comparing the results, we draw the conclusion that it’s more accurate to describe the scanning optical surface with parabolic cylinder equation.

The flame image monitoring technology based on CCD is a research focus of the boiler combustion control, and the key of this method is to obtain a clear and reliable flame color image. In response to this demand, wavelet analysis and dual-threshold method are used to enhance flame color image. Firstly, the color image is decomposed into three two-dimensional RGB signals. Secondly, the RGB signals are processed through wavelet transform respectively, and those wavelet coefficients are amplified and reduced using dual-threshold method for the purpose of enhancement and denoising. Then, the new RGB signals are reconstructed by those wavelet coefficients. At last, the new RGB signals are regrouped to get enhanced flame color image. Experimental results show that, after processed by this method, the image has a strong sense of hierarchy, detail of image is outstanding and the distinction of the combustion zone of the flame is obvious. Moreover, measurement precision of two-dimensional temperature field is improved by 0.62% and the absolute error is reduced by 5.3℃. This is helpful to improve the performance of the boiler combustion control system.

Locally Linear Embedding (LLE) manifold learning algorithm needs to calculate the neighbor points of each image based on Euclidean distance. But this method represents only the straight line distance between two points and does not necessarily reflect the actual distribution relationship of the image data sets in the high dimensional space. In order tosolve this problem, an approach based on the fusion data between angle and Euclidean distance of images is proposed to calculate the neighbor points of LLE and to classify data. This method uses the fusion data between angle and Euclidean distance of images to measure the adjacent relations of image data points and find k neighbor points, which can achieve more effective local reconstruction to extract the distinguishing features. Finally, the nearest neighbor classifier with angle of images is used to classify the image data. Experiments on KSC and Indian Pine database show that the overall accuracy of this proposed algorithm is improved by 1.54%~6.91% compared with LLE algorithm.

Super-resolution reconstruction plays an important role in reconstructing the image details and improving the visual perception. In the most of the conventional learning-based super-resolution, prior knowledge of the input image itself or natural images database is used to solve the super-resolution problem, so the quality of reconstructed images can be further improved. To reach this goal, the information of the image itself and the natural images database are combined. Firstly, the self-similarities across different image scales can be exploited to construct an image pyramid, and the high resolution image is reconstructed only by the input. After that, we learn a dictionary from natural image patches and reconstruct the initial reconstruction one, which is regarded as the input. In the back processing, non-local similarity and iterative back-projection are exploited to further improve the quality. The experiments show that the proposed algorithmachieves better results than other learning-based algorithms in terms of both visual perception and peak signal-to-noise ratio.

The image point feature stability under the Opposite-Movement (OM) and inference conditions is a significant problem in the field of Optoelectronic information. The feature stability problem was researched under the condition when the high speed OM existed between the camera and the scene. First, the principle of the OM blur was analyzed in detail. And then, the stability variation law under the OM blur was analyzed by the experiment and the sifting method of the Harris corner under the OM blur was presented. For the OM blur, the radiant distance of the feature to image center has an important implication on the feature stability. On the basis of traditional Harris corner uncertainty method of Hessian matrix, the radiant distance weighted measure was proposed. The experiment results revealed that, under OM blur condition, the sifting method proposed here could exclude the most unstable feature, and furthermore significantly improve the repeatability of the active feature set.

Aiming at the problem that Fourier descriptor has not enough information for shape contour’s fine details, an approach combining Fourier descriptor and corner feature into one kind of improved shape contour descriptor was proposed. Difference of Gaussians (DOG) operator of scale space or difference of inverse Fourier transform were used to detect corner in order to improve the calculation speed, and the method of minimum cycle distance was used to compare corner sequence in order to makes the matching more robust. Experimental results verify that the approach greatly improves accuracy rate compared with Fourier descriptor approach and improves calculation speed nearly 10 times compared with Curvature Scale Space (CSS) approach.

A novel matching method based on Robust Nonlinear Projective Nonnegative Matrix Factorization (RNPNMF) is proposed to find the correspondence among different images containing the same object. We show how the features point-sets can be matched using their common projection space. The contribution can be divided into two parts. Firstly, a robust RNPNMF method is developed to capture the common projection space of the feature point-sets. Secondly, a matching approach is derived from the projections on the common projection space of the feature point-sets. Finally, two experiments are conducted to verify the effectiveness of the proposed method. The experimental results show that compared with the existing method, our method is more effective in matching the feature point-sets and can be generalized well to registration.

The method for converting from line segment form to chain code form is improved to solve the defects of the existing one. In the original method, 6 kinds of segments were proposed according to the different characteristics of the line and converting them to chain code by their types. This works only for the general situations. When the specials such as the target are only one pixel width, the segment won’t belong to any one of above categories, and the program will go error or even blow up. Two new kinds of segments are proposed to make up for the deficiencies of the existing algorithm and the detail process of them is provided. Experimental results show that the new method can be the perfect realization of the binary image contour extraction by converting from line segment form to chain code form.

As the traditional optical encoder has the deficiency of coding complicatedly, high price and difficult to miniaturization, this paper introduces a new coding scheme of single-ring absolute optical encoder, and then illustrates the decoding and absolute-position calculation method of the coding grating ruler image, which is acquired through array image sensor, and processed by using digital image processing technology. The results of both computer simulation and actual measurement show that, the single-ring absolute optical encoder encodes simply, and has high measurement accuracy. And it is not sensitive to the changes in light intensity. The encoder has low installation requirements and a broad application prospect.