Location:Home > Engineering science > Detection Technology and Automation > Research on Moving Object Tracking Algorithm for Intelligent Visual Surveillance

Research on Moving Object Tracking Algorithm for Intelligent Visual Surveillance

Downloads: []
Tutor: WangQingXian
School: Lanzhou Jiaotong University
Course: Detection Technology and Automation
Keywords: Target tracking,Mean Shift,LBP histogram,Kalman filtering
CLC: TP391.41
Type: Master's thesis
Year:  2011
Facebook Google+ Email Gmail Evernote LinkedIn Twitter Addthis

not access Image Error Other errors

Video-based moving target tracking in computer vision, image processing and pattern recognition has long been a very important and active research topic. In recent years, it is mainly used in intelligent video surveillance system. Intelligent video surveillance systems, video tracking research a challenge to accurately track moving targets under different environmental conditions, is the practical application of a pressing problem. Papers for different scenarios moving target tracking problem, moving target tracking algorithm based on Mean Shift to study the core, the main work is as follows: Research video-based target tracking technology, video-based motion tracking algorithm classified and pointed out the advantages and disadvantages of the various target tracking algorithm. Based on the type of target tracking features in video target tracking technology, and also pointed out the advantages and disadvantages of these characteristics have different tracking algorithms. The Mean Shift algorithm and its application in the tracking of moving targets. In visual tracking process typically requires the user selected at the first frame of a video sequence for the target tracking, and create a histogram of the target. Mean Shift algorithm best candidate region Bhattacharyya coefficient, Mean Shift algorithm iterative search target model in subsequent frames, the method showed good performance in tracking, such as real-time target deformation, partial occlusion, robustness. But when the target and the background is too similar to the poor separability between them, the modeling approach is difficult to distinguish between the object and the background, so that the algorithm tracking failure. Mean Shift tracking algorithm, commonly used as a description of the target mode normalized weighted color histogram. A normalized color histogram is a discrete estimation of the probability density distribution of the target color. Since the color histogram describes the statistical characteristics of the target whole, and does not include the spatial information of the target, therefore, the use of the method of sub-block of the target area, and were calculated for each sub-block Bhattacharyya coefficient, and then select the largest sub Bhattacharyya coefficient the center position of the block to calculate the center position of the entire target, and then use the position update other Bhattacharyya coefficient is relatively small the position coordinates of the sub-blocks. The experimental results show that the algorithm can effectively track a moving target with good accuracy and robustness. Framework for Mean Shift tracking algorithm using only a single color characteristics of said track the target, the target of the method to introduce color and texture histogram modeling method instead of a single color histogram modeling method. Traditional Mean Shift tracking algorithm, the lack of a forecast update mechanism to track complex target tracking scene, often leads to the problem of tracking failure, the introduction of the Kalman filter method predicted moving target in the frame occlusion of the approximate location of the point, its true location as the starting point of the Mean Shift iteration, and then use the Mean Shift algorithm in the neighborhood of the point to find the target, you can solve the moving target. Tests prove the algorithm changes in pose, illumination and direction of movement of the moving target to achieve good tracking results, and also have good robustness for occlusion.
Related Dissertations
Last updated
Sponsored Links
Home |About Us| Contact Us| Feedback| Privacy | copyright | Back to top