Abstract:The new concept of optoelectronic intelligent information processing system technique is proposed and technical characteristics are analyzed. The research results. in relevant science and technological fields are described. Finally the trend of development and prospects of application of this technique are discussed.

Abstract:Based on the idea of applying the techniques of hybrid electro-optic real-time pattern recognition to scanning tunnel microscope(STM), a type of high efficiency and good intelligence STM is proposed. The constructing principle of the intelligent STM has been introduced and the results of computer simulating test have been given.

Abstract:A new 3f multiple hybrid opto-electronic roal-time pattern recognition system is introduced. A planar microlens array is used to generate multiple Fourier transforms of the input patterns. The matched filters are binary phase-only filters fabricated by standard VLSI techniques on a fused quartz substrate. This system performs the correlations of the input patterns and many reference images.

Abstract:Based on the techniques of three levels classification and dynamic scene analysis, the prototype multi-target recognition and tracking system can not only segment faint moving targets from slowly varying background, but also track moving multi-target, predict impending occlusion, reacquire and track targets which wander in and out of the field of view. The system has high percentage of correct identification, and performs real-time processing. The processing rate of target classification is 10 frame/second.

Abstract:The image information of an object is extracted by way of imaging. The estimate theory is used to take the optimum estimate value of the state of the object. The self-adaptive method is used to regulate the parameters of the tracking filter, thus attaining the goal of intelligent tracking. The simulation results indicate that good tracking effect can be achieved by using the above method.

Abstract:In this paper, based on the content addressing process of associative memory, the optimization method of binary associative model is proposed by using a neural network. The error tolerance of the optimum model is analyzed statistically. It is shown that the performance of the optimum model is improved.

Abstract:A pipeline filtering algorithm for digital images ispresented. The procedure is nearly real-time. The VDI (from camera); FB (frame buffer) and a VPI (output of pipeline) can be selected as the input of the filter. The filter function is changed by using the tablet and the spacial coefficients are estimated. Thereafter, the filtering is completed by the ALU pipeline processing.

Abstract:This paper introduces the design of a system for multi-region segmentation and feature extraction in infrared images in real-time. The system has solved the hard problem of segmenting multiple regions in real-time, and adopts the approach of operations in parallel for acquiring region's features. The test has shown that the design of the system is feasible and successful.

Abstract:The invariant moments of an image, which exhibit three inherent properties, i. e. the invariance under image shifting, rotating and scaling, are described. And a hybrid optical digital processor consisting of a mask, two lenses, a CCD detector and a microcomputer, is set up for extracting the invariant moments of the image. The experimental results show that the measured invariant moment values of six English printing letters are indeed unchanged under letters shifting and rotating. In consequence the processor can be applied to pattern recognition for image shifting and rotating.

Abstract:This paper presents three superresolution imaging methods, including the linear prediction data extrapolation DFT (LPDEDFT), the dynamic optimization linear least squares (DOLLS), and the Hopfield neural network nonlinear least squares (HNNNLS). The live data of a metalized scale model B-52 aircraft mounted on a rotating platform in a microwave anechoic chamber and a flying Boeing-727 aircraft were processed. The imaging results indicate that, compared to the conventional Fourier method, either higher resolution for the same effective bandwidth of transmitted signals and total rotation angle in imaging or equal-quahty images from smaller bandwidth and total rotation angle may be obtained by these superresolution approaches. Moreover, these methods are compared in respect of their resolution capability and computational complexity.

Abstract:It is argued that it is the receptive field of visual neurons that enables people to maintain approximately color constancy despite variation in the spectral power distribution of the ambient light and despite variation in the color of nearby objects. A lightness algorithm based on the model function of receptive field of human visual neurons is proposed and is proved to be a general color-constant algorithm as well as an adequate model of human color constancy performance. Problems about lack of color constancy of colorimeters are briefly discussed.

Abstract:Light-excited C_(60) is studied by electron-lattice coupling tight-binding model. When one or two electrons are excited from HOMO to LUMO, the lowest energy state is ring bipolaron with the following characteristics: (1) The bond structure is layers with seven different bonds. In comparison with the dimerized state, bond variation is like that of ring polaron: to be the largest in equator, nearly unchanged in poles. (2) Two energy levels get into the gap and form the self-trapping electronic bound states which localize in the equator area. (3) C_(60)'s symmetry reduces from I_h to D_(5d). (4) There exist nonlifiear exciton and 4 NMR lines.