Based On The Mean Shift Image Tracking Technology And Realization

Image tracking is a hot spot currently in the domain of computer vision. Meanwhile, image tracking has wide applications in visual surverillance, image compression,3-D reconstruction and military affairs. Through the study of object tracking algorithms, one of the algorithms has been applied to the DM642 in order to shake off the fetters of Personal Computer. This thesis tries to get insights on some key issues in image tracking; these issues include updating of object's model and prediction of object'motion, etc.The main contributions of this thesis can be concluded as follows:1. The research of the common object tracking algrothms, including moving object detection and tracking algorithm under the stop background, relative matching tracking algorithm, particle filter based tracking algorithm and Kalman filter based tracking algorithm.2. Mean Shift based tracking algorithm is chosen as the basic algorithm of the thesis's tracking algorithm. Becaused of the disadvantage of the basic mean shift algorithm, some improving techniques have been applied to the algorithm, including adaptive updating of object model, estimation of object's motion and the processing strategy when object occludes, etc.3. The above tracking algorithm has been achieved basing on the research of the hardware of DM642. Through the requirement analysis, software frame designing and the system optimizing, the tracking system has been comleted in this thesis. Some comparisons among aboving algorithms have been made in theory analysis and experiment simulation area. Then the mean shift algorithm is chosen as the main tracking algorithm in this thesis. Kalman filter is chosen as the estimation algorithm of object's motion. At the end, the algorithm has been achevied based the development environment of DM642. Through the result of the experiment, the system has got good effect. And this has laid a sound foundation for engineering applications.