The Applicantion Of Mean Shift Object Tracking Algorithm In Video Sequences

Moving object tracking in video sequences is a basic and important problem of visual analysis, advance behavior understanding and movement coding, in computer vision applications. Its main work is focuse on the video image sequence of moving target detection, extraction, identification and tracking; obtain the motion parameters of moving targets, such as position, velocity, acceleration, etc., as well as the target track. Thereby further processing and analysis are using to achieve understanding the behavior of moving targets in order to complete higher level tasks. Computer-vision-based visual target tracking has attracted a large number of researchers involved in many research institutions. Many abroad research institutions also list video-based object tracking as an important research field, and have made a lot of research results. But the general sense of the tracking technology is still far from mature; to develop a truly robust and practical application of object tracking system still have a large number of issues need to be resolved.Thesis focus on the complexity background scene moving object tracing problem, take video-based Moving object Tracking Algorithm as study object, in-depth study of the mean-shift algorithm in the complex environment of the object tracking applications, the main works are as follows:Thesis has study the current video-based object tracking technology, and discussed the object representation, discuss the research of video-based target tracking technology that the characteristics of target tracking methods and selection criteria; make classification for the video-based Moving Target Tracking Algorithm and points out the advantages and disadvantages of the target tracking algorithm.Thesis focuses on moving target tracking problem, an analysis of video-based tracking of moving targets in the working principle and the main component modules. Focuses on the principles of digital image and video image pre-processing technology, computer vision in the existing theory on the basis of the framework proposed in this paper used the theoretical framework for computer vision. Introduced the mean-shift tracking algorithm the basic idea of the advantages and disadvantages of the algorithm is analyzed.Since the defect of mean-shift algorithm can not track the fast motion moving target, we use the mean-shift algorithm and Kalman filter with a combination of moving target tracking algorithm, make the mean-shift algorithm has been improved. Kalman filter using the operational forecast track moving targets, and then on this basis the use of mean-shift algorithm can track the goal precisely. The simulation results show that the algorithm can effectively track moving targets and a very good solution to the problem of tracking error accumulation.the traditional mean-shift object tracking algorithm, lack a template update mechanism, using a fixed template of the target tracking in complex scenes, often leads to failure. Kalman filter based on the mean-shift group template update algorithm, the characteristic of target in the feature space of the probability information as a template to design a filter bank. Through the Filter analysis of residuals, the development of the corresponding template update strategy, testing to prove the algorithm to ensure that changes in the posture of moving targets, under the light of changes in tracking the effect of block also has a good robustness, but also to ensure more good real-time.Finally, the full text of the research work are summarized and pointed out that future work in the direction of further study.