A STUDY ON VEHICLE EXTRACTION AND TRACKING USING STEREO

This paper describes a new method for extracting traffic information using a pair of stereo road images to track the vehicles. In the extraction process, the vehicle outline is obtained using a binary region growing technique applied to a disparity map based on multi- resolution stereo matching. In order to determine the flow of vehicles, a Kalman filter tracking algorithm is applied to the extracted vehicle. Results from experiments indicate that the proposed method led to significant improvements in the recognition rate of vehicles over the conventional frame difference and background elimination methods.