Vehicle Segmentation by Edge Classification Method and the S-T MRF Model

In this paper, we propose a tracking algorithm, which is based on the collaboration of the S-T MRF model and a dedicated segmentation algorithm. Although the S-T MRF model was designed to be robust against occlusion, it regards vehicles that move in parallel occluding each other from the beginning to the end of the traffic images as a single region. In order to compensate such a defect of S-T MRF, we have developed a dedicated segmentation algorithm which decides boundaries of vehicles contained in such a single region by referring to the difference of edge patterns among the vehicles. By the experiments using traffic video from three different angles at different locations, our method was proved to be very successful.