Embedded Video Tracking Algorithm

In this dissertation, automatic video detection and tracking algorithms of moving targets have been presented based on accomplishing general design of intelligent-embedded video tracking system and selection of suitable software and hardware. Study of automatic detection and tracking algorithms is the most important content in this dissertation.Motion detection and tracking algorithms have been accomplished through a large number of experiments and plentiful analysis in the scene backgrounds that can be indoor or outdoor, simple or complicated. And algorithms have been emulated and testified by MATLAB. Contribution in this dissertation as follows:General design of intelligent- embedded video tracking system has been accomplished. It includes development flow, selection of suitable software and hardware, framework design of software and hardware, compartmentalization of function modules, and so on.In this dissertation, Otsu and mathematical morphology methods for segmentation of difference images have been adopted using differences of three-frame sequential images. These methods need less memory, implement easily and fast, and are convenient for hardware implemention. They have good effects on motion detection of one or multi-moving objects.Mot ion tracking algori thm is made up of motion estimation of KALMAN filter algorithm and feature matching algorithm. KALMAN filter can be self-modified by its remain and plus. Motion-model parameters can be modified using real-track of tracked object. This tracking algorithm is easy and fast. And It adapts to hardware implemention. Further more, it need not to save former datum and need less memory. So it accords with requirement of embedded- video tracking system.