Research On Motion Target Detection And Tracking Algorithm

Along with the wide application of computer vision in various fields, the research of target recognition and tracking has very important theoretical value and practical significance. Each researcher is trying to improve the accuracy, real-time and robustness of target tracking, the reliable target tracking algorithm can get reliable target information, and then lay a good foundation for the subsequent environmental monitoring, scene analysis.Firstly, this paper studies the method of moving target detection, the main target detection method, because the frame difference method is easy to be interfered by noise, the real-time performance of optical flow method is low, so the target detection method based on Gaussian mixture model is selected, and the principle of Gaussian mixture model is analyzed, the good detection results are verified by a set of comparative experiments. In view the fast tracking of Meanshift algorithm and Camshift algorithm, so in this paper, the method of moving target tracking Camshift and Meanshift to do a detailed introduction, but Meanshift algorithm is unable to adjust the target tracking framework, Camshift can be used to adjust the tracking frame and update the target template, which is based on the Meanshift algorithm,but it is easy to be influenced by the background environment. The accuracy of target modeling is affected, and the target is lost.In order to improve the accuracy and stability of target tracking, this paper improves the accuracy and stability of target tracking in two cases. When moving target and background color contrast is low, the target tracking will be affected greatly, So this paper proposes a new method to improve the tracking effect by two dimensional joint information, When the target is in a very complex environment, there always many same colors with background, affect the target tracking, so this paper proposes a new method of Camshift tracking based on histogram saliency., the target model is updated by comparing the target and the background histogram, these two methods are better to establish the target model to reduce the background information of the interference in order to achieve the results of a stable tracking targetThe results of the simulation experiments show that the two methods proposed in this paper can effectively improve the accuracy of the target model and achieve the stable tracking of the moving target.