Research On Image Matching System

With the rapid development of modern society,image matching is used in the video tracking,target detection,modern military and medical diagnosis widely.Knowledge of image matching covers computer artificial intelligence,machine learning,digital image processing,computer morphology and machine vision.The features of images based on image matching is a hot area,and the performance of image matching algorithms based on features is mainly determined by feature space,similarity measure,transformation space and search strategy.In some common matching algorithms,distance is often used as a measure of similartity,hausdorff distance is applied widely.It can be used to judge the matching of two point sets,and has good anti-noise property and adaptability to image shielding.Based on the original hausdorff distance,theory of the algorithm is analyzed in-depth,and several better hausdorff distances are taken out and their advantages and disadvantages are analyzed.Then a new method of distance measurement is proposed.Finally,the matching algorithm is tested on DM642 and used in engineering applications.This article studies in the following aspects:1)The method of image matching is summarized,and some feature extraction methods and image matching based on features are introduced.2)The theory of hausdorff distance is introduced.Its advantages and disadvantages are also analyzed.In order to solve the problem that traditional hausdorff matching measures only use the spatial position between points set,an improved weighted hausdorff matching algorithm is proposed,which combines the direction difference response of the corner point and the edge point.The theory of the algorithm and the corresponding weights are expressed.The algorithm is tested on five groups of images.Experiment results show that the algorithm can achieve a higher correct rate in matching.3)The chip of TMS320DM642 has the characteristics of high-speed operation,the characteristics is made full use of the image matching system based on TMS320DM642.The matched image collected by CCD camera and the matched image with the template image are matched by the algorithm.The system not only can meet the requirements of fast matching,but also matching location is more accurate.