| Moving object detection and tracking in image sequences is the main research field in computer vision, which has been widely applied in military, medicine and scientific research etc. The accuracy and stability of tracking effect depend on the design of algorithms to a great extent. It is still a hot point to research an accurate and robust algorithm in computer vision.First, as for moving object detection, the thesis mainly studies a few of image difference algorithms, and put forward one background updating method based on many image sequences.Then, the thesis analyses and studies the problem of moving object detection and tracking in image sequences acquired by a mobile camera, and presents a new method of background motion offsets based on block match. The method not only resolves the bad effect of camera moving to object detection, but also the target extracted from real-time images is relatively entire through the method. Through the method combined with general image difference, the thesis presents a new method of moving target tracking based on moving camera. Through the tests of tracing for one person in complex and static background, the result of tracking appears relatively good, but the speed of tracking needs to be faster.Finally, as for the moving object tracking algorithms based on template matching, the paper mainly studies the SSDA and template updating rules. In order to continue to enhance the speed of SSDA in matching and searching a target object, the thesis presents an improved method based on Mallat wavelet transform. As to the problem of object tracking failure which is owing to bad template updating , the paper present another method of moving object tracking based on difference and SSDA. It combines with the merits of image difference method. When the tracing for target is failure, it can automatically extract moving target template and recapture the object. According to many tests of tacking one person, the method is proved to be feasible. |
PDF Full Text Request