视频监控中移动目标检测跟踪-信号与信息处理专业论文.docxVIP

AbstractVideo surveillance technology has a wide range of applications in security, financial, transportation, electric power and other departments, and with the development of society, it is more and more important to develop intelligent monitoring system .Since intelligent video surveillance systems has been focus on a very long time ,this paper choose the core of intelligent monitoring system which is object detection and object tracking as research contents, providing a solid foundation for future intelligent monitoring.In the front part of this thesis, there are introduces of the stage of development of video surveillance, after that ,I introduces the composition and principles of network video surveillance, analyzing the difficulties of detection and tracking, propose new methods to improve performances of related system. Object detection is the premise of object tracking ,we often use optical flow, background difference ,the frame difference, to detect moving objects ,however ,those methods can not finish the job of object tracking in a good performances. I propose a fast object detection algorithm on the bases of background difference and frame difference. This algorithm able to achieve real-time target detection under complex background.I propose an method to improve Mean-shift tracking algorithm based on segmentation algorithms. The segmentation method is in order to solve the problems of object occlusion. And experiments show segment method can indeed solve the occlusion problem, But it is too computationally intensive causing Mean-shift algorithm behave badly in real-time performance. So in this case I propose a fusion mechanism to solve this problem. Set all those segmentation in different values. This method solve above problems successfully for both real-time and robustness performance.Key words: Video surveillance, object detection，object track ,Mean-Shift algorithm, segmentation Mean-shift目录第一章 绪论 ....................................................