Based Video Traffic Vehicle Detection Technology

With the progressing of technology level in recent years, intelligent transportation systems as a new mode of traffic management which has been beginning to gradually improve traffic management. With less human intervention, technical updates, and efficient management methods, it has been the direction of the future transportation system development.Based on the videos of the Traffic vehicle detection technology is one of the most important technology of the intelligent transportation systems. Its research can extend the Traffic vehicle detection technology and improve the deficiencies of the traditional vehicle detection technology. Thus, the paper focuses on the following aspects.First, in the extraction of the target vehicle, the paper analyzes some commonly used methods which have Multi-frame averaging method, continuous frame difference method, optical flow method, Gaussian mixture modeling approach. Further programming of the above four algorithms, simulation results show that Gaussian mixture modeling algorithm is better and more conducive to the transport vehicle segmentation. In order to improve the effect of the extraction of the target vehicle image, in this paper, we research filtering, morphological method and removed the shadow method for reducing noise.Secondly, in the aspect of object vehicle detection, Camshift algorithm has been used to track the target vehicle in tracking of the target vehicle firstly. Simulation results found that it will be miscarriage of justice or lose tracking goals when there is only a little color contrast between the target vehicle and the background. In view of this situation, the paper presents improved Camshift algorithm. The improved algorithm replaces the tone reverse projection image of the target vehicle with the target vehicle binary image obtained by the Gaussian mixture modeling method. Improved algorithm simulation results show that improved Camshift algorithm can accurately track the target vehicle when the color contrast between the target vehicle and the background is very small. On this basis, the article introduced the Kalman filter prediction algorithm to solve the problem of vehicle overlap or objects blocking. Experimental results show that through adding the Kalman filter, we can resolve the interference problem which caused by vehicles overlap or objects block, and always able to accurately track the target vehicle.Finally, in the article, we provide three kinds of methods of transport vehicles detection which have vehicle counting, vehicle detection and vehicle retrograde judgment method. We make the simulation experiments of vehicle counting. The data which get from above simulation experiments is accurate. Thus show that the image information extracted by the above approach is effective and feasible method.