Vehicle Behavior Analysis Using Target Motion Trajectories

In this paper, a real-time vehicle behavior analysis system is presented, which can be used in traffic jams and under complex weather conditions. In recent years, many works based on background estimation and foreground extraction for traffic event detection have been reported. In these studies, the vehicle images need to be accurately segmented, although uneven illumination, shadows, and vehicle overlapping are difficult to handle. The main contribution of this paper is to make a point tracking system for vehicle behavior analysis without a difficult image segmentation procedure. In the proposed system, feature points are extracted using an improved Moravec algorithm. A specially designed template is used to track the feature points through the image sequences. Then, trajectories of feature points can be obtained, whereas unqualified track trajectories are removed using decision rules. Finally, the vehicle behavior analysis algorithms are applied on the track trajectories for traffic event detection. The proposed system has been used widely by Chinese highway management departments. The application performances show that the newly developed system and its algorithms are robust enough for vehicle behavior analysis under complex weather conditions.