Video Based Vehicle Detection And Tracking Algorithms Research

The detection and tracking of the moving objects is one of the most important branches in ITS, which combins advanced technologies and reserch achievements in image processing recognitions, artificial intelligence, automatic control and other relative fields. It has broadly applied in video surveillance, robots navigation, video transmission, video retrieval, medical image analysis and other fields, so this subject has important theoretical significance and wide practical value.The aim to this paper is to design video-based vehicle in the highway detection and tracking system, which can be used in large area, multiple objects and complex enviroment, and exclude the noise of the system and track the object effectively. The innovation and the major works of this paper include:1.Studying the detection of the moving objects. Since the background is stationary in most of video surveillance, this paper proposes the moving objects detecting algorithm which uses background subtraction in the moving region. In this algorithm, firstly, the background is reconstructed and renewed which uses the algorithm of stationary background based on dynamic update, this background can update adaptively with the change of the illumination and the environment. Having good background, then, the moving objects can be segmented by background subtraction. The obtained difference image need a series of treatment. So firstly, the connected regions are searched by the method of mathematical morphology; secondly, no car noises are excluded by three determining conditions: area, length and width; lastly, the including and combining operation will be used. Experiments show that this algorithm reduces the influence of the noises in the subtraction image to the detection result of moving objects, and excludes no car noise effectively.2.Studying the tracking of the mving objects. This paper proposes the moving objects tracking algorithm based on the variation characteristics of vehicles moving space. In the ITS system, since the reason of the visual field, the coordinate of the actual road is different from the coordinate of the image, the allusion model of the object and the image will be firstly established, which can proiect the coordinate of the actual road to the image, so the basis will be existed which is used to distinguish the area, length and width of the vehicle in every frame. Since vehicles are rigid bodies, in the movement process of vehicles, the characteristics of the region of vehicle head stock do not change, then, the paper will use dynamic, multi-scale wavelet transformation to locating width and length of vehicles. In the width location, the shadow will be removed, and the abscissa of vehicle head feature point will be located too; In the length location, no car noise will be excluded further, and the ordinate of vehicle head feature point will be located too. Afterward, the track of the target will be formed when the vehicle feature points of the same target in every frame are continued. The experiments show that this algorithm is highly efficient real time, fast and accurate.3.In the analysis phase of experiments results, three samples which were collected in different time, different weather and different emvironment, and the standard which be used to distinguish that vehicles are tracked successfully or not. The results show that the rate of the real time tracking successfully is more than 92%.