Research On The Algorithm Of Object Selection Based On LEGION And Its Implement For Color Image

Object selection is an important but hard problem for image process, pattern recognition, artificial intelligence and it's a fundamental process of computer vision technology. The pattern selection and recognition of an image is depending on the result of segmentation. With the research of the mechanism of human visual becomes advance and the rapid development of computer technology, it provides new ways to solve the problem of object selection of an image. Method with artificial neural networks or genetic algorithms promotes the image segmentation to intelligent way. D.L.Wang and D.Terman proposed Locally Excitatory Global Inhibitory Oscillator Networks (LEGION), and they use this model for segmentation and object selection of an image.This paper studies the model of LEGION selection networks based on oscillator correlation and two-stage object selection networks. Then, the author simulates the object selection networks using LEGION on the computer, studies the influence of parameters to the networks, simulates the LEGION object selection algorithm, adjusts the algorithm parameters to observe the result of selection, and achieve a property result. Then the author compared the result with other common object selection algorithm.Currently, LEGION is only used to select the object of binary image and gray level image, however, in the real world, the color image contains more information and is more widely used. The author introduces an object selection method using LEGION based on RGB model and HSV model to select the object of color image. Then, the author simulates the object selection networks using LEGION on the computer and analyzes the result of simulation.