Study Of Object Detection And Tracking In Transportation Video Detection System

With the development of communication, transportation video information processing is becoming more and more important. And video-based motion detection technology is one of the most important methods for transportation video information processing. How to detect the objects to collect traffic information quickly and accurately is the key point of video detection technology study. Based on summarizing and analyzing of current target detection and tracking method, this paper is focusing on moving object detection and tracking technology under static background. Some key technologies on object detection, such as modeling and updating under static background, object segmentation, object tracking are intensive studied and discussed, some of the related algorithms are improved and implemented. And the results of experiments are analyzed. On the end, a transportation detection system is implemented based on these research results. The following work has been down in the paper:1,After Analyzing popular theories and algorithms of static background-based moving object detection, this paper summarizes all algorithms' features. Based on these features, the paper presents an improved object detection method which combines frame difference and background difference;2,With Deeply studying on background extraction and analysis of current background extraction, the paper presentes a new background extraction method that is improved from auto-modeling using adaptive background subtraction; Then a background updating algorithm with weight factor is presented after analyzing the theory of background updating;3,This paper studies basic theory and method of object segmentation; focusing on threshold selection method of object segmentation, the paper describes threshold value calculation method using most Otsu method adaption calculation in details, and then describes theory and method of separating target pixels and background pixels by this method. Particularly, the paper describes the process of eliminating noise by expansion, corrosion method, and combining neighborhood using region filling to obtain the ideal split target area.4,Based on analyzing the theory of MeanShift target algorithm for color features with-out parameter estimation, the paper implements the algorithm, compare and conclude the advantage and disadvantage of the experiments results and algorithm. Then the paper studies extended Kalman filter tracking algorithm and analyzes its feature. The target tracking and extrapolation is implemented.5,This paper presents a tracking method based on blob and describes the thought and frame of blob tracking method based on OpenVC architecture. Base on that frame, a detection model is defined and implemented in the paper;6,This paper uses C++ and OpenVC technology to develop a transportation detection system based on video images sequences on Windows XP with the help of Visual studio 6.0. Then the paper develops a transportation parameters analyzing and displaying subsystem. The system has been passed the tests, the results of tests indicates that algorithm meets reliability and real-time requirements.