Multi vision detection is an image processing technology which captures images of the goal from different angles with several CCD sensors, and obtains the information of the image after matching, segmentation and minus. The multi vision detection technology can obtain more information of the image than monocular vision, which reduces the complexity of follow-up treatment and improves the measurement accuracy of detection systems as well as the anti-interference ability and measurement efficiency. However, in order to obtain images with different information, it requires better image and lighting systems than monocular vision system. Based on the requirements of the experiment, the lighting system of multi vision detection is designed, and the experiment platform is set up. It is confirmed that the type of light source, the structure of lighting, the optical properties of detected objects, and the background characteristics are important factors for visual inspection technology, and the research provides the experimental basis for the promotion.

In the passive estimation of mobile target track, the stationary ground platform only provides angular information but no distance information. So, the target flying track must be estimated solely with the existing angular information. A new algorithm, based on uniform acceleration and constant altitude flying mode hypothesis, is presented to estimate the mobile target track according to the measured angular information. A simulation model was built to test the performance of the algorithm. The experimental results show that the estimation algorithm can reduce the angular error to milli-radian or less.

According to the analysis of laser-pulse encoding modes, a method to decode laser-pulse series based on autocorrelation matrix statistic is put forward. The received laser-pulse series are used for constructing first order difference autocorrelation matrix and second order difference autocorrelation matrix. The elements of the two autocorrelation matrixes are counted in histogram mode. The laser-pulse series encoding mode can be classified by the characteristic of the two histograms, the time parameter of the pulse can be calculated out. The validity of the algorithm is tested by emulating experiment. Different modes of encoding laser-pulse are generated, and the missing probability and false probability are set, then the laser-pulse series are decoded by the algorithm. The emulating experiment shows that the algorithm has high precision, good real-time and anti-interference performance.

The image received on the image surface of detector will rotate regularly with the platform rotation. Based on the system’s basic frame, the optical transfer model was analyzed and one handy method was suggested to offset the image rotation. This method makes use of the actual result of the image processor and needn’t use optical way or electronic way to offset image rotation. This mathematic method is very easy, increases the system processing speed and reduces the system’s complexity. It is practicability.

Based on the vector theory of optical reflection, nonlinear scanning traces characteristics of the two-dimensional pointing mirror in the scanning mode in the given area is analyzed. The expression of scanning trace and image rotation angle and the method of non-linear scanning trace modification are given. Without eliminating the image rotation, this mode can be used in point object detecting system, such as IR detecting system.

In order to study the effects atmosphere transmission on the laser beam intensity distribution at sea, a model of laser beam intensity distribution on reaching face is established based on the example of Hermits-Gaussian (H-G)s. According to the generalized Huygens-Fresnel principle, the peculiarity of laser beam intensity distribution is simulated under the different condition such as refractive index of atmospheric structure constant and transmission pace and laser wavelength. The result indicated: if the parameters of effect and beam is confirmed，the peculiarity of laser beam intensity distribution in far field can be determined.

To meet the requirement of tracing large quantities of targets real time, an approximate multisonsor joint probabilistic data association(JPDA) algorithm were discussed after studying the joint probabilistic data association (JPDA) methods. This method adopts the conception of approximate cluster which uses approximate cluster matrix to get the association matrix and breaks the association matrix by dividing a large association matrix into some small association matrixes of non-intersection each other. This method reduces the counts of observed and echo and the complexity and quantities of the algorithm. In addition, two kinds of algorithms, parallel and serial, of multisensor joint probabilistic association algorithms were analyzed and compared.

The status and problems of infrared and radar stealth technology in present are reviewed briefly. New infrared radar compatible stealthy materials being developed in foreign countries, such as doped semiconductor oxide, electric conduction polymer, retinyl schiff base salt, hollow electric particulate coated with metal and nanometer absorbing materials are discussed in detail. Finally, stealth mechanisms and development status of new infrared radar composite stealth coating technology are described.

Based on introducing the form and the principle of the system for large-aperture collimator, the indoor emendation measurement of excessive optical axis parallelism for photo-electricity measuring equipment was done using the measurement principle. The optical axis parallelism emendation for TV/IR/Laser was implemented through fabricating standard laser target and collecting laser spot of the laser target by sensitive paper, so that the optical axis non-parallelism of TV/IR and Laser meets the technical index requirement, the good effect is realized.

The structures and working operations of SOA-assisted Sagnac XOR gate, all-optical MZI XOR gate and the SOA-assisted bidirectional data input XOR gate were analyzed, then the three XOR’s performance and the complexity of the system were compared. Based on it, a modified SOA-MZI-based all-optical XOR logic gate was proposed. The feasibility of the program was analyzed. Finally, the optical principle and mathematical model of the program were discussed. Values analysis shows that the program can effectively achieve the XOR function, and appropriately adjust the parameters, the system performance will be superior to the existing XOR gate.

A fiber grating structure with cladding radius as sinusoidal function along the grating length is introduced. The transmission properties of the this type grating was studies on the basis of the transfer matrix theory. Compared with linearly chirped fiber grating, the fiber grating exhibits a Box spectrum with steep edges and high reflectivity when tension is applied to it. With the tension increasing, the center wave length in the grating reflecting spectrum will move toward the long wave direction and the reflecting bandwidth is greatly widened.

For the CCD serial star images, if there is a high bright star in the field of view, the CCD readout circuit voltage can be over high and the same line or column gray value of the CCD image is on the high side, a light line traveling across the star centre is formed in the image. As the bright line exhibits great similarity to the motion of space targets, and disturbs greatly target detection and identification performance, it must be eliminated. Based on the analysis of the star-sky image, an automatic detection algorithm of bright line in star-sky images is put forward. With the isotropy property of star imaging and auto-registration character of Fourier transform, the algorithm can automatically detect and eliminate the bright line both in space domain and frequency domain. Experiments of 3 groups of star-sky images (60 frames per group) show that the algorithm can automatically detect bright line with high accuracy.

The Random vibration experiment is a common simulation test in the engineering experiment, while applying the vibrate quantity class, it will probably outrun vibration set extreme limit and result in breakage. Through enumerating a few key index signs of vibration pedestal directly influencing the random vibration experiment, and according to the vibration condition of product, the full thrust, the largest displacement and the maximum acceleration required to random vibration experiment were estimated. By contrasting with the vibration set extreme limit index signs, the vibration set was tested to be capable of carrying on that experiment.

Automatic detection and recognition of targets bases on optical images is a hot point of research. Fractal theory provides a new technical way. Because the difference of fractal dimension between artificial target and natural background is obvious, an electro-optical target detection method based on fractal dimension is proposed. Experiment proved that the given results obtained by the targets detecting method are stable and superior to the edge detection method and the threshold segment method, and it is an effective detection method.

The single target image information characteristics are analyzed, three methods of processing three-dimensional image information of the targets from a single theodolite are put forward, it has an important practical application value.

Through discussing the video signal processing method for the traditional CCD observation camera, a newly signal processing circuit solution centered on VSP2212-high integrated special CCD video processing chip was designed. All the associated debugging of hardware and software were accomplished．The results of simulation and experiments show that the system is with high integration, low consumption, stable operation and low noise, making a better basis for the designed CCD observation camera system to develop into a light and compact system．

How to control the movement of the laser head accurately is the core problem of laser marking system. Interpolation and calibration algorithm, which directly affect the quality and results of marking, are the most important part of laser marking system. A marking motion control system is introduced, which makes use of DSP-BF533 as a hardware platform and adopts the optimization of linear interpolation algorithm and the calibration algorithm. The system successfully obtains a satisfactory output of vector graphics based on the marking precision, therefore, enhances the processing precision of marking.

A new method to enhance the edge of infrared image was given，which is based on the adaptive threshold and the corner limited in an appointed sector．The SUSAN operator is introduced, and an improved method to get the adaptive threshold and limit the corner in an appointed sector is presented．The clear and continuous edges of infrared image were got in practical application．Experimental results show that the new method can enhance effectively the edge of infrared image and the result is satisfied．

Aiming at the problems of multi-target tracking in video sequence, a multi-target tracking algorithm based on mean-shift and particle filter is proposed, which resolves the problem of target shielding by adopting the distance threshold and similar degree threshold among the targets in combination with Mean-shift algorithm and particle filter tracking algorithm. Experiment results show that the multi-target tracking algorithm can reduce the tracking time while ensuring the effects of tracking multi-target, so that the capability of the designed tracking system is improved .

The smoke effects are analyzed , such as visible light shielding; the effects contrast of the visual luminance of background with the visual luminance of the target ，the dispersion/ reflection of the laser. The model of smoke effects for transfers of lightis discussed. By analyzing the theory of the model, the working relationship between the smoke effects for transfers of light and the smoke properties are indicated.

Based on the theory of expert systems, simulating thoughts process of experts, gathering domain knowledge of all experts and according to the problem characteristics of application and management system, an expert systems model of malfunction intelligent diagnose has been built up. In the system, the method of knowledge representation is genetic rule, the inference engine combines uncertain inference and rule inference. In order to deal with the uncertain knowledge, an uncertain inference model based on confirmation theory is established. This model can auto-gets and modify knowledge in knowledge base, achieve a better real-time characteristics, through interacting between human and machine as well as self-learning function.