Study Of The Neural Computational Model Of Moving Object Contour Extraction Based On Retina Physiological Mechanisms

The retina physiological functions, corresponding computing mechanism, and it's neural networks based mathematical models, have been already the research focus of the vision and its application. This dissertation gives a retina mathematical model with which image edge and contour detection can be enhance by using moving information, after doing a intensive study on the retina physiological functions and computing mechanism of motion detection and edge detection and gets a series of simulation results by proposed model in edge detection application.The researches of this dissertation are as follows:1,A depth survey of retina physiological functions, it's mathematical models features and the receptive field of kinds of retinal cells, including the function of lateral inhibition among retinal cells in edge and contour extraction of the mechanism, the function of the bipolar cells in image edge detection mechanisms, the function of the amacrine cells in motion detection and the function of OMS ganglions in motion detection. And then, map a neural network topology consists of luminance channel and motion detection channel by rods/cones,horizontal,bipolar,amacrine and ganglion, further, including the characteristics of the network layers of cells and receptive field structure.2,Giving a detailed analysis about the interaction between luminance channel and motion channel detection in several possible methods in the retina. Giving a mathematical model about how the motion information can enhance edge detecting base on the interaction between luminance channel and motion channel detection. This mathematical model uses the motion information enhance edge and contour detecting. Lastly, give a series of edge extraction simulation results by the proposed model.Finally, this dissertation summarized the main research results and pointed out the direction for further research and exploration.