Research On Key Techniques Of Motion Target Detection And Tracking In Static

Moving target detection and tracking is an important and forefront subject in computer vision, it involves in many advanced science and technology fields, such as pattern recognition, artificial intelligence, image processing etc, is widely applied in the military, intelligent transportation, video surveillance and so on. So the research on moving object detection and tracking has important significance and practical value. At present, the detection algorithm for dim moving target and illumination changes, in the light of the moving target object occlusion and pose variation under the condition of the tracking algorithm of moving target detection and tracking is the unsolved key problems.Based on the above key technology problem, the paper is focus on detecting and tracking the moving object under static scene. Some of the traditional algorithm is improved, and the algorithm is realized by programmingOn the research of moving object detection, this paper integrates the Gaussian mixture background model and three frame differencing to detect motion object. Firstly,the paper introduces three traditional algorithms, including frame subtraction, optical flow and background subtraction. All the advantages and disadvantages and scope of application of those algorithms are analyzed. Then this paper introduces some traditional algorithms of background modeling in background subtraction, including the median method, the moving average method, W4 method, single gauss mixture model and mixed gauss model. Because of the three frames difference method can’t detect the object integrality, and the Gaussian mixture background model is sensitive to the change of scene, so we combine the two methods to detect motion object. According to the different combination forms, the paper gives three different kinds of new algorithms, and analyzes and summaries the effect of detection and processing time of eight algorithms.Experimental results show that the new algorithm is better than other algorithms, and lay a good foundation for the subsequent moving target tracking.On the research of object tracking, this paper presents a new moving target tracking method that combined with the new detection method and TLD algorithm. The current popular Tracking-Learning-Detection target tracking algorithm includes tracking module,detection module and study module. It can achieve success in tracking a single goal long time, but when the illumination changes and the targets are subject to pose, scale, its tracking effect is poor. In view of the above problems, the paper combines with the new detection method and TLD algorithm. This method can solve the target tracking problem such as in the case of severe occlusion and pose variation, and enhance the robustness and accuracy of tracking algorithm.