Research Of Object Tracking And Video Surveillance Technology Based On MPEG-2 Video Sequence

Video object tracking is a new research subject that develops with the improvement of digital video technology. The results of object tracking reflect on the main information contained in the video stream, thus this technology is widly used in such fields as video transmission, content based video searching, and digital video monitor sistyem. This paper reports on the motion object tracking technology and its application in the digital video monitor system.First, this paper address the problem of using motion models based on MPEG-2 motion vectors in classifying and tracking objects when prior information of the object is not available. A novel algorithm based on MPEG-2 video stream is put forward which combines advantages of Shio and Sklansky algorithm and global light stream algorithm. Unlike previous applications of colour and complex shape in object tracking, our method is designed to track targets commonly found in indoor visual surveillance, and to meet the real time need. The input of the algorithm is MPEG-2 video stream captured by a stationary camera. Then the algorithm tracks moving objects in the stream without decoding it. The motion vectors in MPEG-2 video stream are directly used as the feature of moving objects. After a grouping and data association operation, motion parameters and structures can be aquried.And then, a design plan of digital video surveillance systems based above-motioned algorithm is proposed in this paper. The framework consists of monitor camera, teminal server part and network part. It is a set of LAN-based system that has such functions as romotecontrolling, motion detection and so on. This paper describes the design details, including the demanding analysis, overall design, module design, function design and interface design.Last, we have a test on the algorithm and its PC application with MPEG-2 sequenses. According to the experimental result, we analyze the algorithm's advantages and disadvantages. The results show that the algorithm provides a robust visual tracking system in MPEG-2 video stream which is capable of performing accurately and consistently within a real world visual surveillance arena. The system can successfully track multiple moving objects independently and have theability of the approach to maintain trajectories in the presence of occlusions and background clutter. It also meets real-time need in the moving objects tracking application.