A tracking algorithm is proposed 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 as to be robust against occlusion, it regards vehicles that move in parallel occluding each other from begin to end of the traffic images as a single region. In order to compensate such a defect of S-T MRF, a dedicated segmentation algorithm was developed which decides boundaries of vehicles contained in such a single region by referring to the difference of edge patterns among the vehicles. This method is called the Barcode method. In addition, a method is proposed to obtain precise trajectories of vehicles against such serious occlusions. In this method, a motion vector of a vehicle will be added to the initially obtained mass center of a vehicle. And the height and length of the vehicle scaled by geometrical factors will be considered to obtain a precise trajectory of tail or head of the vehicles. These methods enable the precise tracking even in such seriously occluded situations. By experiments using traffic video of three different angles at different locations, the method proved very successful. For the covering abstract see ITRD E134653.
Abstract