Road Traffic Surveillance with Multiple Views

Recently, since the car is often used as the movement means, the traffic is increasing every year. ITS (Intelligent Transport Systems) which provides drivers with plural information for smooth and safety drive has been required. As one of them, several studies have been proposed for the road traffic surveillance in the image processing. T h e traditional methods have almost observed the traffic flow from one view~oint with one camera or stereo-cameras. In this case, it is difficult t o detect the part of one vehicle occluded by another one with single view. Our proposed method for the road traffic surveillance extracts vehicles with multiple views. Extracted motions in the different viewpoint are complementary t o each other, and compared by using the estimated camera parameters of which cameras are placed freely by road managers, and so on. As a result, the occlusion of vehicles is detected. Our implemented method has been tried t o real traffic scenes and successfully detected the individual vehicle .