Real-time traffic information extraction based on compressed video with interframe motion vector

Extraction of traffic information from image or video sequence is a hot research topic in intelligent transportation system and computer vision. A real-time traffic information extraction method based on compressed video with interframe motion vectors for speed, density and flow detection, was proposed for extraction of traffic information under fixed camera setting and well-defined environment. The motion vectors are first separated from the compressed video stream, and then filtered to eliminate incorrect and noisy vectors using the well-defined environmental knowledge. By applying the projective transform and using the filtered motion vectors, speed can be calculated from motion vector statistics, density can be estimated by using the motion vector occupancy, and flow can be detected by using the combination of speed and density. The embodiment of a prototype system for sky camera traffic monitoring by the MPEG video has been implemented, and experimental results prove the effectiveness of the proposed method.