Research On Target Recognition Algorithm Based On Contour Segment Weighting

Shape matching is an indispensable step in target recognition system,and object recognition based on contour segments is a very promising and necessary research direction.In this paper,the contour extraction algorithm,the contour segment acquisition algorithm,the credibility evaluation of contour segment and the weighted similarity matching algorithm of contour segment are studied.The main research contents are as follows:A contour extraction algorithm is proposed.In order to eliminate the influence of noise and other factors in the transmission of color image,gray image enhancement operation is carried out to improve the image quality.After graying,the image is smoothed and denoised first,and then sharpened and enhanced to make the image edge(the gray level changes greatly)become more vivid.Then the binary processing is used to reduce the amount of data.Finally,the object contour is extracted by edge detection and morphological processing.Simulation results show that the algorithm can effectively extract the target contour and retain the effective features.A target recognition algorithm based on contour feature fragments is proposed.In order to obtain effective segmentation,the contour segmentation algorithm based on discrete curve evolution and the segmentation optimization algorithm based on curvature are studied firstly.By comparing the experimental results,it is concluded that the contour segmentation can better express the target features by considering the contour curvature.Then,the effective segments are matched by shape context,and finally the optimal matching is performed The recognition results are obtained.The simulation results show that the more curved contour segments can get more accurate classification results in shape matching.A target recognition algorithm based on weighted contour segments is proposed.Aiming at the low recognition rate of traditional target recognition algorithm in contour matching,a weighted similarity matching is proposed to improve the recognition rate.Firstly,two parameters of locality and importance are proposed to evaluate the credibility of segmentation,and then weighted similarity matching isproposed,which combines the similarity between segments and its credibility,and finally gets the weighted similarity to measure the recognition accuracy.It can effectively reflect the relationship between local feature similarity and global feature similarity,and get the best classification result.The simulation results show that the weighted similarity effectively improves the target recognition rate,and achieves significant results in partial occlusion target recognition.The interface of the target recognition system is designed,and the above algorithms are visualized.Four menu pages,contour extraction,contour segmentation,shape context matching and weighted similarity matching,are designed through the GUI platform of MATLAB software,which makes the operation more flexible and convenient,and the usability and portability of the algorithm used are higher.