Detection And Tracking Algorithm Research Based On Vehicle Traffic Video

In recent years, Intelligent Transportation System(ITS) plays an important role in dealing with transportation problems. It is an inevitable process as well as one of research hotspot of ITS to take real-time detection and vehicle tracking. The thesis is a sampling analysis of video streams in motorways to describe the solution of vehicle tracking and tracking algorithm in details. Then according to the established technologies of vehicle detection and tracking, the author made the improvements in related algorithm and provide an improvement project in such application scenarios. The thesis mainly includes the following steps:(1) Research on algorithm of background upgrading methodIn algorithm of background, the author employs the motion object detection technique based on the requirement of real-time detection and background difference. According to the introduction of the commonly used background modeling method, the author comparative analyses advantages and disadvantages of it and comes up with an adaptive background updating method. The method uses the way ofweighted accumulative frame difference to realize the automatic updating of the background. The advantages of this method are less calculation, faster background updating rate and relatively short background and initialization time.(2) The removal of shadow in the day and night lightsDuring the day, the sun will form a shadow on the ground, and the shadow of these motions will have a great influence on the accurate extraction of the contour of the moving vehicle. If a car travels at night, duing to the dipped beam irradiation in the ground forms a bright spot, the spot will result in a bad influence on the detection. Thus the thesis effectively solves the problem of distinguishing day and night time. At the same time, in different application scenarios, using different algorithms to solve the problem of shadow and spot as well.(3) Vehicle detection in congestion and non-congestion situationsIn the phenomenon of road congestion, the thesis uses the virtual loop technology to detect moving vehicles, and in non-congestion, detected by wire technology. The thesis compares the advantages and disadvantages of the virtual coil and the detection line technology, and points out the respective application scenarios. And practicable methods are put forward here to realize the automatically switch of tracking algorithm in the road congestion and non-congestion situations.(4) Research on vehicle tracking algorithmThe author analyses the previous vehicle tracking technology, and designing a tracking algorithm based on characteristics of the vehicle. The target contour of the moving vehicle has been extracted after binarization processing, so the effect of the illumination can be negligible. This thesis addresses the problem of tracking failure duing to the overlapping of moving target vehicles in the tracking process. But there exists an information loss from the 3D projection to the 2D plane, the image of the vehicle pixel distance is not equivalent to the real distance, and without solution to the problem of projection correction. Therefore, the estimation of vehicle speed and position is still short of accuracy and need to be further studied.All above, the thesis designed a complete system of vehicle detection and tracking based on transportation video, and works out the research results. The results show that the system have good real-time and accuracy in vehicle detection to achieve the initial tracking effect and obtaining satisfactory results.