The Research On Tracking Method Of Moving Vehicles Based On Computer Vision

With the continuous application of video surveillance system in the field of traffic, motor vehicle tracking based on computer vision has become the foundation of intelligent traffic monitoring system, which s is used for vehicle speed detection and information acquisition. In practical tracking applications, the diversity and complexity of vehicle targets are the inevitable limiting factors in the process of target tracking. Therefore, a video target tracking algorithm with high stability and robustness is needed to deal with these problems.In order to achieve real-time and stable tracking of moving vehicles on the road, the target detection and tracking method of moving vehicle is studied in this paper. Aiming at the complexity of the environment, and the specific work is as follows:1) According to the research background of the subject, the research status of target detection and tracking is analyzed. According to the development of video tracking, and the difficulty of moving object tracking technology is proposed.2) In the image pre-processing stage, this paper ultimately selected median filter to remove salt and pepper noise by the analysis of comparative experiments to deal with salt and pepper noise existing in the visual image. First of all, three commonly used moving target detection algorithm are analyzed. Then using the mixed Gauss model to describe the background. If the background information which is produced by the light and the camera shake is strongly changed, the measure factor should be introduced to adapt to the change of the environment, as the same time, the background should be updated in real time to realize the adaptive switching between the background subtraction method and the frame difference method. The experimental results show that the algorithm can resist the interference of environment changes, and it can achieve the accurate detection of moving objects.3) Aiming at the background interference which may appear on the highway, the Mean Shift algorithm is designed for the target tracking. First of all, selecting the target rectangle region as the initial search window by manually, as the same time, the object model is built based on the kernel histogram of RGB, which is to find the optimal center window location. Experimental results show that the proposed algorithm can overcome the interference of light, shadow, partial occlusion and other background factors, and achieve real-time tracking of the target.4) Aiming at the lack of the ability of the mean shift algorithm in the target tracking for fast moving vehicles and occlusion handling, The algorithm design of moving vehicle target tracking based on particle filter is presented by using the theoretical knowledge of particle filter, Through the operation process of the random sampling particle and the updating weight, the stable tracking of the interested moving vehicle is realized, and the problem of target tracking is solved better, and the experimental simulation results verify the effectiveness of the algorithm.5)To solve the problem of the complexity of particle filter, a moving target tracking algorithm based on Shift Mean and particle filter algorithm is proposed. The position of the particle distribution is close to the true position of the target by the clustering algorithm and the position estimation of the observation model. According to the illumination change, shadow interference and shielding phenomenon, the experiment results show that the algorithm is real-time and robust, and it can achieve a more accurate and stable positioning and tracking for the moving vehicles.