The Research On Chaotic Neural Network With Its Application On Computer Vision

Computer vision is a comprehensive and challenging subject. So far, plentiful fruits have been got in that research. The neuron operating speed is only on millisecond in human visual neural system, which is very awkward to computer. However, depending on these neurons, human visual neural system realizes operating high graphics visual dada real time and transmitting them high speed, which go so far ahead computer visual system. Then artificial neural network is born for simulating human neural system.Chaotic neuron is a complex system of computing simulation for biological neuron, while dynamics is a good means to describe it. Being based upon MATLAB, The chaotic neuron dynamic system had been portrayed visually by script and plot in the paper with programming, calculating and simulating. The dynamic analysis for chaotic neuron grounding on thus had indicated it that the input item including external information has fateful effect on the interior status of chaotic neuron dynamics. As the input item varying, chaotic neuron status transforms between chaos and periods, which should be used to control the interior status of chaotic neuron.Chaotic simulated annealing neural network (CSAN) structured by chaotic neuron is a kind of powerful tools on optimizing. But the research on search mechanism of this neural network doesn't catch our eyes deeply, which has cumbered its ability on application for more. As a rule, making use of chaotic simulated annealing neural network (CSAN) to optimizing has to proceed two progresses: Firstly it carries through "coarse search" up to a superior value in the network's annealing process; then it switches to grads "minute search" based upon the value to approach the best value. Basing on the analysis of the single neuron's dynamics characteristic, the article successfully controlled its chaotic action by a implication for period doubling bifurcations feedback, went deep into the process with CSAN solving TSP problem, then found out the "coarse search" mechanism of CSAN: In the annealing process, the self-feedback item serves as the chaotic power, and the coupling item which products continuous feedback serves as the perturbation, which can find out a attractor corresponding to the best or superior value under the stabilization of period doubling bifurcations.Founding on the groundwork of search mechanism to CSAN, the article constructs a kind of computer visual system to simulate human eyes' movement mechanism, which realizes the human eyes' basic functions such as orientation and identification and presents the simulating results.