Support Vector Machine-based method is pre sented to recognize human faces. Compared with traditional methods, such as Nearest Neighbor Rules, Euclidian Distance, Mahalanobis Distance and Neural Networks, this method achieves higher recognition rate. It can classify complicated patterns and overcome disadvantages of overfitting in traditional methods. The process of this methed is as follows: pre-processing the human face images first, then using primary component analysis (PCA) to extract and select the appropriate features of human faces, training multiple SVMs by the face feature vectors, and using the trained SVMs to classify human faces at last. The performance of SVMs are also compared. It is concluded by experiments that the second-order polynomial SVM has better performance than other SVMs on human face recognition.

A Synergetic Association Classifier (SAC) based on Synergetic theory is presented, in which a dynamic system is constructed. And by the evolving of the system and finally reaching a steady state, objects could be classified accurately. In this paper, a project which recognizes images of vehicle-license-plate using SAC is also devised. The segmented characters are identified by SAC. The experiments show that the results of the recognition of vehicle-license-plates are satisfying. The algorithm is easier for training and has good capability of association. It is robust for image distortion caused by noise and visual distortion.

According to the hue, motion and geometric feature s of human faces, a method integrated hue analysis and motion detection was raised to remove non-face regions. The remaining regions are called candidate human f ace regions. Then the candidate regions are verified by the geometric features o f human face and a trained Support Vector Machine, so human face regions are det ected. The advantage of this method is that the total process is divided into tw o independent modules, the first module is to minimize the omission rate, while the second module is to minimize false rate. This kind of system structure can p romote both detection rate and response speed greatly.

An improved KNN method is presented in this paper. Features are selected by the principle of surface smoothing and according to the target value distribution situation of the points found by traditional KNN metho d. This improved KNN method is based on target value, it can avoid the influence of subjective classification compared to the traditional KNN method based on classification.

Due to the complexity of underwater sound field an d huge amount of memory requirement, the distribution of underwater sound field ne arly couldn't be fully described with PC computer. In this paper, the visualizat ion of 3-D slices of sound field based on pseudo-color image by the theory of computer graphics as well as the IDL language is realized. The simulation result truly reveals the preconceived distribution of the sound field.

Data visualization is an important research area in computer science. In this paper, data visualization technologies including data pre-treatment, mapping and displaying are briefly introduced, and its applicati ons in medicine, weather forecasting, engineering and data mining are presented also.

In this paper, the method of setting up a MIS (Man ag ement Information System):" Database Management System for Data Fusion Engineeri ng", which is based on common gateway interface(CGI) and communicate through Win Sock, is introduced. The structure, constitution and application of the DMSDFE a re described. The difficulties of manage data with data fusion technology are il lustrated. The differences between DMSDFE and other MIS are presented.The furth er development of such system is put forward.

Many algorithms have been put forward for binary image registration. A new algorithm based on point mapping is presented in this paper. The method is robust to find the most matches among point mapping between real-time image and model image. Comparing this algorithm with a fast algorithm b ased on Hausdorff Distance and an algorithm based on cross correlation in time d omain it is proved that this algorithm is the same robust as the other two algor ithms by theory analysis and experiments, but this algorithm is more efficient in computation, and of practical value.

In the scene matching performance evaluation system, the feature about the similarity between the reference image and real-time ima ge has the same importance as the statistic image features in image feature set. Some correlation plane statistic features are defined, and the experiment resul t manifests the relations between the features and matching performance. These f eatures and the statistic image features can constitute the image feature set.

The turning kiln group is an important equipment in metallurgic industry. Its temperature state that is reflected by kiln's skin dir ectly influences the production efficiency and the product's quality. In this pa per, an infrared imaging system (IRIS) that is used in real time group temperatu re measurement for turning kiln's skin is described. First of all, the principle of the system is presented, then the system design and the temperature standard ization are detail introduced. Some results of the system application are also d escribed. At last the practical function of the infrared imaging system is discussed.

In this paper, improving methods of purifying raw el ements and melting technology of As-Se-Te glass fibers in 2～12μm wave band w ere introduced. Effect of the fiber loss, packing fracting K and surface ref lection on the transmittance was analyzed respectively. The loss of As-Se-Te g lass fiber measured by FTIR spectrometer was obviously lower. By using the As-S e-Te glass fibers, 100×100 pixels 2m length coherent fiber bundles were made. While coupling one end of the bundles to the target area of the IR TV camera thr ough the lens system, the other end was aimed at a 3mm diameter hole at 5m away. This small hole had been homogeneously injected by ideal radiator. The thermal image was delivered through the boundles and was clearly observed on the TV scre en. The results show that As-Se-Te IR coherent bundles are one of the best des ign for delivery of thermal images in special circumstances

In this paper, radiation characteristics of target /b ackground and atmospheric transmittance are analysed. By analysing the result of the test, infrared focal plane array sensor working within 3～5μm wave band is considered to be appropriate for detecting and tracking conventional fast movin g object. Based on this, the detecting distance for 122mm rocket using infrared focal plane array sensor within 3～5μm wave band and the precision of two infra red measuring system intersecting measurement are discussed.

In this paper, basic theory and key technology of tr ansient laser holographic interferometry are described. Some experimental result s are given. Recent research and developed instruments for holographic interfero metry are introduced. The four parts in this paper are following: 1) describing the physical process and mathematical model on off-axis holography;and 2) key technology of transient holographic interferometry;3) introducing typical equip ments for measurement;4) discussing several problems.

Image acquisition and processing are essential tec hn ologies in laser imaging radar system. Designing procedure and composition of th e image acquisition and processing based on Master/Slave (M/S) computer system a re presented. Interrupt functions of M/S computer system are utilized to perform frame-real-time acquisition, processing and displaying of laser image. A spec ial DMA controller is designed to co-occupied two identical dual-port SRAMs wi th slave computer ADSP21020, and successive image buffering, acquisition and pro cessing are implemented by exchanging enable rights of SRAMs synchronously. Fina lly, the experiment result analysis is given.

In the laser images,the target underwater is inunderated by the strong laser beam and the strong laser beam has single direction.Because of that ,in this paper,mathmatical morphological operators are used to process the image to eliminate strong beam disturbance to target detection.Then, target image can be obtained and triple correlation analysis is proceeded on it .After that the target is detected accurately .Experiment proves that it is a fast efficient and robust target detection method.

In this article the laser diffraction technology is extended to a new application field to measure the thickness of 1.5-meter-wide soft graphite roll. The broad graphite roll in high quality has wide application, but its production process must have corresponding real-time and nondestruct ive measuring and observation system. Using low-power laser as illuminant, a se t of applicable metrology scheme that could measure both the thickness and the u niformity of graphite roll is advanced -- relative method is used to measure the thickness and multi-point measurement method is used to measure the uniformity . Some discussion on the selection of the related important parameters are also given here, and some modified measures concluded from experiments are proposed. What's more, the real-time processing on diffraction stripe is studied.

Fabricating nano-Si by LICVD, we discover: there is a low threshold of laser intensity and a high threshold of SiH 4 flow speed, a nd they are positively correlated to maintain the high temperature for SiH 4 to split. In order to obtain small and even nano-Si particles, the laser intensit y must be large and the SiH 4 flow speed must be fast correspondingly to increa se the nucleation rate of nano-Si particles, and reduce the numbers of Si atoms absorbed by each nano-Si nucleus, and shorten the growth period of ea ch nano-Si nucleus. When nano-Si is annealed for dehydrogeniza tion after it is prepared, its IR absorption spectrum is changed. The positionintensities and shapes of four characteristic absorption spectrum bands are changed respectively,becaus enanoSisurface is large and oxidized so that its constituent is changed.In order to weaken such oxidation,nanoSi should be annealed in Ar but not in the air,and the annealing temperature should be lower than300℃.

The technique of laser interference is important f or resisting laser-guided weapons and self-defending. In this paper, the passive interferences of smoke、aerosol and balloon, and active interferences of deceit ?suppressing and near exploding fuse are studied. The techniques of laser inter ference which are used in truck-load?plane-load and warship-load weaponry system are analysed. The developing status of this technique is summarized.

In this paper, temperature increment of thermal resistor irradiated by laser is calculated, from which the relation between its temp erature increment and laser power is studied. It indicates that the output signa l of thermal resistor detector are almost linear with laser power when it is lower than the breakdown threshold of the themal resistor. The curve of this relati on is obtained by the experiment. The results of calculation and experiment are compared.

An image representation method based on arbitrary shaped regions is proposed in this paper, and the motion estimation of the image sequence is executed according to this representation. The initial frame is smoothed while edge of the image is preserved. The smoothing algorithm is the modification version of Alvarez's method. The smoothed frame is segmented by the watershed method. According to the marker image, the image is stored in the form of region adjacency graph. To solve the problem of the over-segmentation, the criterions of region intensity and edge strength and region size are combined. The affine transformation is used as motion model for each region, and the nonlinear least squares method is used for the optimization. Compared with the method based on pixel, the result shows that the motion vectors produced by this method are more consistent and robust and the PSNR is improved.