Key Technology Research And System Development Of Target Detection And Tracking In The Video Surveillance System

To get real-time information of the field people need not get to the location for the application of the video surveillance system. With the processing capacity improvement of computer technology and the development of computer pattern recognition, it could be possible to realize automatic understanding the video context by add pattern recognition technology in the video surveillance system. The video surveillance system with pattern recognition ability will improve work efficiency and has broad market prospect.This thesis firstly introduces the technology of third generation video surveillance system and object detect and tracking, and then gives a deep analysis on the principle of object tracking and detecting. Finally, some key methods that used to improve algorithm performance and real-time processing performance are proposed. A video surveillance system with intelligent analysis function was developed. The thesis does exploratory researches main include two sides: key methods research and system development.Some key methods researches main about how to improve algorithm performance and real-time processing performance. For improve real-time processing performance:①A method to get information blocks quickly is presented, with which could get the targets'information includes size and location. Some methods to improve algorithm performance are also based on this method.②A method that used to process binary image fast is presented when need to use binary image by sequential access or random access.③A method that used to locate target and match target fast is presented. For improve algorithm performance:①A method to correct and de-noise processing the targets detected result with the use of target blocks is presented. With which could improve the detection precision with low process cost.②A new algorithm fusion method is presented, from the character that the advantages and disadvantages of background subtraction and frame difference are complementary in a way, to update the background of background subtraction algorithm with the result of frame difference is presented. In this method, the background change could be judge and the error-detection could be eliminated adaptively. As a result, the adaptability and the robustness of the algorithm is be improved.③parameters and modeling method of the codebook are optimized, the problem of codebook application in real-time system is be solved effectively. Researches on system development:①A returnable program module is developed that can be used to detect moving targets and track target the user defined. The practicality of the module is be optimized to satisfy common demand, and the module can self-adaptive eliminate kinds of error-detection, and can process multi-channel video. Both visible image and infrared image could use it.②A practical intelligence video surveillance system has developed: except the common function of the video surveillance system, this system could set surveillance region in all kinds of shape, when there are targets in surveillance region, the system could play sounds defined by user, like camera number, targets amount, and target location. The system could detect all moving targets and tracking a target located by user. Other function like record video is also designed in the system.