A Suspect Vehicle Tracking System Based on Video

Video surveillance systems are widely used in security field. The large amounts of video data make it time-consuming for criminal investigation. Traditional researches conducted on visual analysis are extremely challenging, due to the complexity of scenes, such as large variation in viewing angle and multiple moving objects. This paper approaches these problems by proposing a video based suspect vehicle tracking system. We used a road modelling based on spatial clustering to obtain colour road information. In order to improve the accuracy of target matching, we fit the 3-D model with region of interest (ROI) to the 2-D target images, and compute the histogram in ROI as the feature for target matching. Experiment shows that system can provide accurate result for suspect vehicle tracking.