Vehicle Violations Detection Method In Video Surveillance

Since reform and opening up, the development of economy has leaded to the increasing of private cars. Although, the growth of the road mileage, the existing road traffic resources are still difficult to support the current traffic pressure. Moreover, air pollution caused by automobile exhaust and traffic congestion may be inconvenient to the people. Intelligent Transportation system(ITS) provides a new way of thinking for sloving traffic issue. As the hot field of ITS, vehicle violation detection method has far-reaching significance in improving and alleviating the traffic situation.This paper focuse on the way of vehicle violations detected in the ITS including vehicle detection, vehicle tracking and achieve recognition and detection of vehicle violations according to the tracking parameters, and the main work lists:Firstly, we study the vehicle detection. Above all, three kinds of background modeling algorithms are compared and analyzed, and an efficient adaptive background updating strategy is proposed. Then, based on the target detection algorithm, such as optical flow, frame difference, background difference, a fast improved background difference method is proposed to get the foreground. From this foreground, target vehicle is separated by the connected region labeling algorithm. At last, under two different traffic scenes, the vehicle detection experiment is conducted and good results are obtained.Secondly, we research the vehicle tracking. Above all, based on the Meanshift algorithm, Camshift algorithm and Kalman filter algorithm, a multi-target tracking algorithm is proposed. Then, we put forward the corresponding solutions while vehicle is blocked and the color of vehicle is similar with background during tracking At last, the experiment results show that the algorithm has good robustness and real-time performance under three different traffic scenes.Thirdly, we study vehicle speed, traffic flow statistics and the vehicle violations detection. Above all, vehicle speed and traffic flow statistics are achieved by calculating the speed and quantity of the vehicle through the area of interest. Then, for the detection of vehicle compact line and illegal retrograde, the recognition of vehicle behavior is achieved by analyzing and comparing the extracted parameters of vehicle trajectory and lane line. For the detection of vehicle parking violations and illegal turning, the vehicle violations are detected through the change trends of vehicle motion trajectory parameters in the area of interest which has been set up.