Pattern Recognition Based On Pulse-coupled Neural Networks

This thesis investigates the applications of PCNN in pattern recognition. Firstly, the paper, combining the characteristic of synchronous pulse bursts and inhibition with the modified PCNN model, presents a way of finding the foveation points in the images adaptively and effectively, and simulates the human vision system. Secondly, PCNN is extended to PCNNs, based on the properties of information couple and transmission, an algorithm that is used to fuse images of the same target got by several sensors to an image is presented to simulate the human vision system. Thirdly, combining the properties of synchronous pulse bursts, capture, and transmission and competition of waves, the paper presents two ways of classification, one is an algorithm based on the properties of neuron to capture and inhibit to classify the data taking on any complex unlinear distribution robustly, the other is based on the restricted distance and modified of the former to remove the influence of inferior samples in classification; Finally, based on the accumulative difference pictures, and the forming and transmission of PCNN wave, selecting and controlling the direction of autowave by connecting the neighbouring neurons selectively, the paper presents a way to simulate the tracks of moving object and detect the moving direction. Thus, we not only reveal the inherent ability of PCNN, but also explore the applications of PCNN in pattern recognition.