This paper presents a vehicle occlusion identification system based on the perceptive characters of a roadway. To tracking the vehicles, two stages are devised, initial parameter setting stage and occlusion handling stage. In the initial stage, a background extracting method is adopted to obtain the first clean background. Then, a road detection algorithm is used for finding the vanishing point. After the lane dividing line detection, the perceptive roadway models are established in each lane. These roadway parameters are fed into the occlusion handling stage for analyzing the vehicle occlusion situations. Due to the different occlusion cases, several novel segmentation methods are proposed for improving the accurate of the tracking vehicles. Experiment results are shown the feasibility of the proposed methods.