Based on the feature analysis of multi-target in the long distance, a new multi-target recognition algorithm based on conformability degree is introduced in the cursory searching process and the precision searching process. In the cursory searching process, all possible targets are quickly searched with the round morphology template in the binary filtered image. When the cluster analysis of all possible targets based on the distance conformability degree principle is finished, the centroid of all possible targets in the same target is used as the center of a local entropy region. In the precision searching period of targets, the position of maximum entropy value is used as growing seed to implement region growing. In order to decrease computing quantity and improve real-time character of system, based on the entropy conformability degree principle, an improved algorithm of region growing which is the combination of the simple linking algorithm and the sub-region combination algorithm implements single target reconstrction.The simulation results show that the algorithm can perform recognition of five targets quickly and correctly.

Based on Fuzzy Hopfield Network, this paper proposes a new recognition system-Based on Core Fuzzy Hopfield Network System (BC-FHNRS). Firstly, the BC-FHNRS trains the core of special target so that the core can be used to symbolize the special target. Then, the unknown pattern target can be estimated by this core to decide whether it is the special target. This paper gives a method of constructing core and proposes a training arithmetic, and the BC-FHNRS recognition arithmetic is also proposed based on these. The BC-FHNRS holds a few favorable properties--less calculating time, higher recogniting probability and higher rejecting probability. Experiment shows that the BC-FHNRS has a favorable performance.

To resolve the problems in real time detection of moving dim small target, such as system dithering and low signal noise ratio, a new method based on mathematical morphology filtering, the continuity of target's intensity and hypothetical test according to the characters of moving dim small target is presented. High-pass filter based on mathematical morphology can suppress clutters and emphasize the target-like peaks. The continuity of the target's photic intensity can be used to identify target effectively. Hypothetical test has attractive result to resolve the problems of target missing and new target occurring. The algorithm has been realized on the TMS320C6201 digital signal processor (DSP), and the experiment results show that the method of detecting the moving dim point target is efficasious when SNR is about 2

To simplify the 3D color head dataset acquired from 3D color laser scanner, a new vertex merging mesh simplification algorithm based on region segmentation is presented. The algorithm can be divided into two stages--region segmentation and mesh simplification. After region segmentation of 3D color head dataset, vertexes are labeled as region-boundary vertexes, which can only be merged into region-boundary vertexes and inner vertexes. Using region-weighted error metrics, the algorithm generates continuous levels of detail. Results of several experiments show the potential of our method for 3D color head dataset.

Aiming at the shortage of contact test in tradition ATS, a PCB fault diagnosis method based on infrared thermal imaging technology is put forward. The hardware configuration and data process method of infrared diagnosis system are introduced, and how to utilize powerful classification and learn ability of ANN to produce fault diagnosis rule is also discussed.

The necessity of cooling of InSb infrared detector and cooling temperature are described. The reasons why the cooling temperature of InSb focal plane detector is stricter than the single element detector are analyzed. It is shown that when the cooling temperature is 77～90K, the detectivity does not change very much. The single element detector still has higher signal-noise ratio at 137K, but signal-noise ratio of the focal plane detector drops greatly. The test results indicate that the focal plane detector using J-T cooler can achieve high performance.

An novel micromachined uncooled infrared detector based on temperature dependence of drain current of MOSFET is proposed. It utilizes a MOSFET fabricated on SOI substrate as thermosensitive element. An IR resonance cavity formed on passivation layer of MOSFET improves IR(8～14μm) absorptivity and a microbridge fabrication using micromachining technique provides a low thermal conductance. The temperature dependence of drain current of MOSFET, IR detection principle and device's structure are presented. The performance of IR detector based on numerical analysis and finite element analysis (FEA) is estimated that the detectivity (D*) can reach 109～1010cmHz1/2W-1.

The principle of sensing of carbon dioxide using IR radiation is introduced and a novel double-beam optical structure is presented in the paper. Based on the good abilities of RBF network such as nonlinear mapping, learning and generalizing, the mathematic model of IR carbon dioxide sensor is established. A dynamic compensation filter is designed to improve the dynamic characteristic of IR carbon dioxide sensor. A novel IR carbon dioxide sensor with high performance is made. The experimental results show that the above key techniques prompt the accuracy, real-time and reliability of IR carbon dioxide sensor efficiently.

It is desired to build a uniform framework to model on data fusion system. Random sets theory offers a kind of potential means to integrate most of the different aspects of data fusion. The basic theory of random sets is presented and the expression of evidence theory, fuzzy sets and conditional event algebra by random sets is discussed which is used to deal with uncertainty, fuzzy and contingent data in data fusion system.

This paper briefly introduces the sagittal focusing and monochromatizing principle about the beam between 5～25keV. Sagittal focusing monochromatizing principle about beam is based on Bragg law (2dsinθ=nλ) and the monochromic beam is obtained by changing Bragg angle θ. According to the relation between θ and sagittal focusing radius R, the monochromator makes the monochromic and sagittal focusing beam realize. Then the paper explains the principle about crystal bending based on flexible hinge, the control about anticlastic distortion, the structure ensuring the fixing height between the input and output beam and the result on the finite element model about the surface of bending crystal. Finally the paper gives the result experiment.

An approach based on evolution strategies is proposed for image segmentation. Taking full advantage of the global searching capability of evolution strategies, based on gray distribution of image and segmentation strategy of fuzzy clustering, the algorithm proposed implements the procedure of segmentation. To make sure that the algorithm can reach the global optimized resolution, a novel method is proposed to improve the globally researching capability of canonical evolution strategy. The experiments indicate that the algorithm proposed can get very good effect of segmentation.

An image processing method based on the multi-scale morphology which focuses on the structural difference between dim target and background noise is presented. A better filtering effect can be got by choosing proper structure elements and morphological operators. The morphological filter is a new type of nonlinear filter developed from mathematical morphology. Structural information distilled from signal by use of morphological operation is determined by the form of structure element, since morphological operators in essence show the reciprocity between the aggregation of targets and structure element. Because of the mathematical preciseness and the simpleness, there is a particular advantage of mathematical morphology when it is used to describe the configurative characteristic of target image. The useful information and multi-level features from the images can be got by multi-scale analysis using morphological filters through changing the scale parameter of the structure elements.

To improve the ground resolution of remote sensing images with aerospace camera, we have analyzed a patent technique used to earth source satellite Spot-5 of French and presented five kinds of possible special structure of CCD of aerospace camera and their ground image processing projects. Three projects of them are described in detail and their representative experiment results are shown. Our computer simulations demonstrate that this study and projects possess great practice signification for the design of aerospace camera and satellite remote sensing image processing.

A method of computerized tongue image analysis based on image processing for the purpose of digitizing the tongue properties in traditional chinese medical diagnosis is presented. The chromatic information and 2D Gabor wavelet transformation are applied to segment tongue from original image. The statistics method is used to identify the colors of each pixel, which belongs to the tongue substance and coating respectively. Thickness of tongue coating is determined by 2D Gabor wavelet transform coefficient energy (GWTE). The distribution of GWTE and invariant moment algorithm are used to judge the tongue venation. The experiment result shows that methods proposed are effective.

With the development of multimedia technology, the rapidly increasing usage of large image database becomes possible. To carry out its management and retrieval, Content-Based Image Retrieval (CBIR) is an effective method. This paper realizes a CBIR system by extracting the spatial texture features and its algorithm and structure of the system are also introduced. The experiment results are given in the end.

A static hand gesture recognition system for human-robot interaction is proposed and realized. Hand area is segmented based on skin color model, then the hand contour is described by use of Fourier descriptor. LS-SVM is used as classifier, which is effective to small sample and has limited generalized error. A new incremental training algorithm for LS-SVM is proposed, which avoids the time-cost computation for matrix inverse. DAGSVM algorithm is used to combine two-class LS-SVM to realize multi-class gesture recognition. The recognition rate of 26 alphabetical gestures is the highest (93.62%) comparing with multi-layer perception, radial basis function network etc.

A new method of feature selection based on association rules is proposed. Association rules have been used in application such as market analysis to capture relationships among items in large data sets. It is shown that association rules can be adapted to feature selection which are used in image analysis and they can automatically identify the most confident feature. Appropriate texture features are selected with this method to serve image classification and simulative results are discussed.

In order to avoid the unobservable problem of passive tracking for single infrared search and track sensor, an active－passive joint tracking algorithm for a maneuvering target with IRST and Laser Ranger is developed. The current statistical model for tracking a maneuvering target is extended from 1D to 3D according to reference[1]. The state and nonlinear measurement equations for an arbitrary maneuvering target in 3D are educed. The angle measurements from IRST and the range measurements from LR are aligned in time. The correct measurement value is given. An extended adaptive Kalman filtering algorithm is used to simultaneously estimate the position, velocity and acceleration of a maneuvering target in 3D space. The results of simulation show that the presented algorithm is feasible and satisfactory.

Detecting and calculating accuracy of centroid of light spot is an important factor which affects the detecting accuracy of Hartmann wavefront sensor. The influence of window selection on centroid detection error is great. The Prieto's method of progressively reducing window size is analysed, and the template matching method is adopted according to the feature of the light spots of Hartmann sensor. With template matching, the accuracy of centroid detection is improved. The computer simulation of the centroid calculations of spot array shows that the centroid's rms error can be decreased by 53.7%. In addition, the template matching method avoides iterative centroid calculations and reduces computer time consumption.

A novel adaptive image enhancing fusion algorithm which is used to fuse visible and infrared image is presented. At first, the new algorithm performs adaptive enhancement of visible and infrared image. Then, the fused image is constructed based on spatial local energy and local normalized correlation measure. The fusion performance is estimated using entropy. Experimental results of visible and infrared image fusion are given, which show that the proposed method is effective and fused image can provide more complementary information for personnel and target recognition.