Intelligent video surveillance is one of the important embranchment in the field of computer vision. The research of intelligent surveillance technology relates to the field of computer science, automatic, image manipulation, artificial intelligence, pattern analysis. The intelligent video surveillance is the system which could analyze and calculate the video image sequence without manual work and obtain the description of the surveillance scene. In general, the procedures of intelligent video surveillance system include motion object detection, motion object description, motion object tracking, motion object recognition and motion object behavior analysis. In fact, not all of these steps are necessary and available in all kinds of application. It depends on the real scene for application. But motion object detection and motion object tracking are the two necessary steps. This thesis focuses on the motion object detection and motion object tracking.Some pivotal technologies of intelligent video surveillance system are introduced, emphasizing on the motion detection and the motion object tracking technology.Two classical methods including Background Subtraction and Temporal Difference are discussed and verified. Some experimental results are given.The motion object tracking tech, also based on the existing tracking arithmetic, focuses on the theory of MeanShift being applied in real-time tracking of classical arithmetic and goes deep into its principle, as well as its realization of emulator. Some experimental results under different contexts are analyzed.In order to reduce the redundant computation and compensate the image distortion, an adaptive post-processing method using mathematical morphology is used before the motion tracking to increase the reliability of the result of the motion object detection. |