Vehicle Abnormal Behavior Detection System Based On Video

With the rapid increase in the number of motor vehicles in China, the contradictionbetween people, vehicles and road have become increasingly prominent. It is urgent toestablish efficient Intelligent Transportation Systems. As a part of the ITS, vehicleabnormal behavior detection has important significance and practical value. With thetechnology development of image processing and pattern recognition, vehicle abnormalbehavior detection based on video has more and more attention of researchers. On the basisof analysis of the research status, this paper reviews three difficult problems of vehicleabnormal behavior detection based on video. These problems are the establishment ofbackground model, the stability of vehicle tracking and the construction of abnormalbehavior recognition algorithms. In response to the problems, the paper provides a series ofmethods and processes.The real-time update background is established by the average background model.The paper realizes detection of moving vehicles by the process of background subtraction,binarization and mathematical morphology processing. In the process of the differencebinary image, fixed threshold value is not sensitive to the external environment. Therefore,the paper proposes the algorithm of flexible adaptive threshold and verifies itseffectiveness by experiments.Vehicle characteristics of center, the length and width of the circumscribed rectangleare extracted. These characteristics are matched by the method of establishing similarityfunction. On the basis of the feature matching, vehicle tracking is implemented by settingthe ROI and track list. And the trajectories of moving vehicles are extracted according tothe list of parameters.The specific abnormal behavior detection methods of the retrograde, speeding, illegalparking and running red lights have been proposed. The experiments of abnormal behaviordetection are carried out based on different video material, and the results are given. Experimental results show that the methods of vehicle abnormal behavior detectionbased on video have high accuracy and real-time, which satisfies the needs of intelligenttransportation system.