The Study Of Techniques On Detection And Tracking Of The Moving Vehicle In Vision

Digital image processing technology is a new science research field which has more development potential. The detection of traffic information based on video is compared with tradition detection, it has many merits, such as the installation simplicity, maintenance expediently, wide range of detecting, various information and so on. Video traffic already becomes an important research direction in intelligence traffic system.This thesis regards traffic scene picture obtained by camera as the research object. The key techniques of background extraction and shadow removal and vehicle tracking are researched and analyzed. Main contents are included:(1) The optical flow, frame-difference and background-difference are analyzed. An experiment is given to prove that the latter is the best for detecting moving object of traffic video;(2) The algorithm for average of multi-frame and the algorithm for background extraction based-on statistic are realized. Many experiments are taken with different updating coefficient a for consecutive frame-difference. This thesis presents a new algorithm for background extraction, it calculates fast. Compared with other different algorithm, this algorithm has a best background.(3) For the question of image segmentation, this thesis introduces many algorithms in common: algorithm of iterative, algorithm of Otsu, algorithm of minimal error, presents a new algorithm of choosing threshold value: automatic choosing threshold value based-on histogram, if choosing an appropriateβ, it can get a fine effect of segmentation.(4)Making use of shadow attribute, it presents an algorithm of shadow removal based-on statistic. An experiment is given to prove that this algorithm can separate vehicle from the joint vehicle because of shadow.(5)This thesis introduces the algorithm of kalman filter and vehicle matching in common tracking, discusses different questions of vehicle covered.