Move Object Detect And Tracking Under Complex Scene

Computer Vision is an important part in computer science and engineering domain. Because of being widely used in military, medical, security and video surveillance, Computer Vision technical has gotten more and more attention. As an indispensable part in Computer Vision, moving object tracking technical plays a pilot role. Object Tracking includes moving object detection, analysis, distilling, tracking, motion understanding. It is a very complex problem; including many subjects such as image processing, pattern recognition, human intelligence and so on.In this thesis, we present a tracking method based on moving object detection, particle Filter, Mean-Shift algorithm.We make summaries, comparison, and research on the correlative technologies of object tracking and moving object detection at the beginning of this thesis. Then we bring forward a moving object detection method based on three frames difference, using OSTU method to get the threshold image, distill moving objects based on the contour.In chapter four, we analyze the mean-shift algorithm, according its shortage; bring forward a tracking method based on moving object detection, mean-shift template updated by mixture Gaussian model. As motion characteristic of tracked objects can not be used in the tracking algorithm based on features, we import the motion estimation method based on particle filter, which is combined with Mean-Shift for tracking. In chapter seven, we discuss the implementation of the tracking system based on object detection, mean-shift, and particle filter.At last, we carry on comprehensive analysis to the operation result of system and point out the future work.