Research On Template Matching And Probability Data Association In Video Tracking With Kalman Filter

Target tracking is a comprehensive project, it includes computer technology, automatic control technology, image processing technology, artificial intelligence, and other technologies. Target tracking has widely applications in practical applications, such as intelligent surveillance, traffic monitoring, shopping malls monitoring, astronomical observation, medical imaging, and military guidance and so on. With the development of computer technology, the development of target tracking has been from the traditional tracking-based on television or screen, which is supplemented by human-based monitoring system to the intelligent direction which is supplemented by monitor-based human.However, there are many problems to be solved in the applications of target tracking. For example, target tracking in the scene of image noise, light changes and color confusion is a basic and difficult problem, which is also the basement of this thesis.Firstly, this paper studies the theories and methods of the moving target tracking. Secondly, the main target tracking algorithm - Kalman filter and the setting of its parameters are introduced. Lastly, two effective tracking algorithms are proposed base on Kalman filter algorithm, and introduce an algorithm - Probability Data Association, it is an extended algorithm of Kalman Filter. Because of the existing noise and illumination and other factors, it will affect the tracking results when we use the traditional Kalman filter algorithm to track on a video sequence, this paper is to research the deficiency of tracking process above, and an extended Kalman filter algorithm is studied in this paper.Noise interference is a fundamental problem in the tracking field, because of the background noises in the image, it will have several candidate areas in detection process, which leads to the wrong track result or the target miss in a video sequence, the dynamic templates and Kalman filter tracking algorithm is proposed in the paper, the algorithm is simple but effective, and it is suitable to the tracking scene of background interference. The scene of illumination changing is an important factor in noise interference. The scene in the video sequence will be affected by the changing of light, it has great influence on the results of tracking, the static template is proposed to select the appropriate target model, and then adopting Kalman filter to track the target, the experiment results show that the proposed algorithms have better tracking precision and accuracy, which can track the moving target effectively and have better robustness in the scene of the variations of illumination. Data association is the core technology in multi-target tracking, probabilistic data association algorithm is a classic algorithm in data association, and the single target tracking is the most fundamental in multi-target trackin, so the single target tracking is studied by using probabilistic data association algorithm, it lays a foundation for multiple target tracking.