Moving Vehicle Detection And Tracking Based On Video Image

In the recent years, with the increasing of traffic, traffic accidents are bound to happen more frequently although the traffic infrastrure improves. Intelligent Traffic System has been put forward to solve this problem, and become a popular research area rapidly as well. Advanced Vehicle Control System in Intelligent Traffic System uses Computer Vision techniques to detect the surroundings of the vehicle. The dangerous state can be identified by addressing the information. Moreover, it can pass the obtained traffic information to drivers to help them to drive automatically or drive with safety driving assistance, avoiding traffic accidents and decreasing the loss of property.In order to improve traffic safety, it is very important to develop car anti-collision technology. The key technology is vehicle detection and tracking technique in Computer Vision. In order to provide driving assistance information to drivers in Intelligent Traffic System, a system for the purpose of vehicle detection and tracking of moving object and dynamic background is proposed in this paper.Firstly, this paper expounds the research objective and current techniques. Also, it researches relative theoretical knowledge of vehicle detection and tracking, including image preprocessing, image feature representing, target classification and target tracking.Secondly, considering the complex background and vehicles in a video image, an algorithm for vehicle detection and recognition based on feature learning is proposed in this paper. The algorithm learns HOG and LBP features of the training images to train weak classifier by means of Ada Boost algorithm to get cascaded reinforced classifier, then using it to implement target vehicles’ detecting and recognition in video images.Thirdly, this paper contra poses the moving target and dynamic background in video images from in-vehicle camera. In order to track moving vehicles, an algorithm of vehicle tracking based on super-pixel segmentation is proposed. The algorithm uses a mean shift clustering algorithm to cluster the super-pixel and then uses confidence value of clustering and super-pixel to track vehicles in video images.Finally, the vehicle detection based on feature learning and vehicle tracking based on super-pixel segmentation are simulated and analyzed. The simulation results show that the algorithms can successfully detect the moving vehicles in the video images, and improve the accuracy of vehicle detection and timeliness of vehicle tracking.