The Vedio Object Tracking System Based On Active Contour Model

Active Contour Models, also called Snake model method, was firstly given by Kass, Witkin and Terzopoulos announce International Journal of Compute Vision in 1988. This model is extensively applied to detection of the edge, image segmentation and tracking moving object, specially applied in tracking of interested object's contour. Kass also provided a solution of tracing of moving object in image sequence with the Snake model in the first time. The advantage of Snake model is integrating all of the image information, initial value estimation, the features of object's contour and the constraint condition based on knowledge in one characteristic extracting process. With appropriate initialization, the Snake model could converge on the energy minimum state. There is many research on improve Snake model with a sort of respects. Based on known research results, we try to development a tracking algorithm with Snake model with consideration for both tracking effects and efficiency, so that the algorithm can be applied to the platform of DSP Image Processing-TDS642EVM. And then, we designed a video object tracking system based on improved Snake model in DSP system.On this video object tracking system, our research works are listed below:1 .With researching on classical Active Contour Models (Snake) , comparison with B-Spline Snake model and other improved snake models, we choose B-Spline as Snake's description.2.Then, we provide an evolution method using low frequency pass filtering with FFT, so that it can be used in DSP system easily and efficiently. Compare with Level Set Method, FFT low frequency pass filtering gain advantage on tracing efficiency.3.We combine the contour initialization based on different information with B-Spline snake model. The result of experiment, which uses video sequence with condition of stable background and illumination, gains good effect.4. Build the video object tracking system based on DSP both in software and hardware, and design application program of it. With the algorithm using low frequency pass filtering based on B-Spline Snake model.