MODEL-BASED OBJECT TRACKING IN MONOCULAR IMAGE SEQUENCES OF ROAD TRAFFIC SCENES

Moving vehicles are detected and tracked automatically in monocular im age sequences from road traffic scenes recorded by a stationary camera . In order to exploit the a priori knowledge about shape and motion of vehicles in traffic scenes, a parameterized vehicle model is used for an intraframe matching process and a recursive estimator based on a m otion model is used for motion estimation. An interpretation cycle sup ports the intraframe matching process with a state MAP-update step. In itial model hypotheses are generated using an image segmentation compo nent which clusters coherently moving image features into candidate re presentations of images of a moving vehicle. The inclusion of an illum ination model allows taking shadow edges of the vehicle into account d uring the matching process. Only such an elaborate combination of vari ous techniques has enabled us to track vehicles under complex illumina tion conditions and over long (over 400 frames) monocular image sequen ces. Results on various real-world road traffic scenes are presented a nd open problems as well as future work are outlined.