Vehicle Behavior Detection Based On Traffic Video And Implementation

In recent years, with the over-rapid growth of motor vehicles and the high traffic accident rate, it has a great impact on our normal life. Now, in China, the main way to monitor vehicle behavior is to view the surveillance video by man, which cannot be foolproof, and that the video can only be checked later and is unable to prevent accidents. So far, in order to further regulate the vehicle to obey traffic regulations, the intelligent vehicle detection has become a hot topic to discuss and research.In this paper, the vehicle behavior detection based on traffic videos is deeply researched and analyzed, according to actual demand of vehicle behavior detection, the methods of image filtering, image enhancement and shadow segmentation to preprocess video images are studied for the contrast analysis, and try to improve these algorithms; Through analyzing the advantages and disadvantages of several common background modeling methods in simulation experiments, the improved algorithm based on Gaussian mixture background model is proposed; In addition, multi-traffic targets tracking algorithm is introduced based upon improving Cam shift algorithm; By extracting video motion center coordinates of the vehicle, the new method of vehicle’s behavior detection based on its moving track is put forward, and also other feasible methods are proposed to judge the behavior of the most common vehicles.In this article, according to above theoretical research, the overall framework of the vehicle’s behavior detection system is built. Using Matlab development platform, both vehicle’s tracking and detection function and behavior analysis function are able to be operated. Furthermore, the feasibility and practicability of the improved algorithms is validated by experiment. At last, try to analyze and explore that how to judge the popular drunk driving behavior。...