Application to the Traffic Occlusion with Multiple Views

Recently, since the traffic is increasing every year, the problems of traffic jam and accidents, and so on, are increasing , it is required that the ITS (Intelligent Transport Systems) provides the drivers with the information for smooth and safety drive. As one of ITS (Intelligent Transport Systems) techniques, the studies on the traffic flow measurement by image processing have been proposed widely. In the conventional methods, since the traffic flow is almost observed from one viewpoint, thus one vehicle is occluded by another one. It is difficult to extract an oc-cluded vehicle by the image processing. In our proposal method, multiple views are used in order to solve the oc-clusion in the image taken from one viewpoint. On the other hand, the locations of cameras are not always placed at the desirable position for the traffic flow observation system. We propose a method to extract the movement vehicles from the multiple viewpoints which are placed as freely as possible. Even if the occluded vehicle exists in the image taken from one viewpoint, our method solves the occlusion by using the image from another viewpoint. As the pre-processing of the solution of occlusion problems, the movement regions extracted from the same vehicle in each view image is matched by the geometric transformation considered the height of the vehicle based on the obtained camera parameters. 1 Instruction Recently, since the vehicle is often used as the movement means, the traffic is increasing every year. For this reason, the problems of traffic jam and accidents, and so on, are increasing. In order to solve these problems, it is required that the ITS (Intelligent Transport Systems) provides the drivers with the information for smooth and safety drive. As one of ITS techniques, the studies of the traffic flow measurement by image processing have been proposed widely. In the conventional methods, since they are almost observed the traffic flows from one viewpoint with one camera or stereo-camera, one vehicle is occluded by another vehicle. Thus, it is difficult to detect the parts of one vehicle occluded by another vehicle with a single viewpoint. The method [1][2] to solve the occlusion problems has been proposed with a single view by S. Kamijo, et al. This method, however, can not be adapted to the case that vehicles are always overlapped each other in the image sequences. It is, needless to say, impossible to detect the vehicle …