This paper proposes a system for grasping the traffic status on roads in real time using monitored images. A conventional method for detecting objects in a time-varying image uses subtraction between an input and a background image (background difference). This background difference technique consists of two steps: object extraction and background updating. In order to achieve accurate background updating, it is necessary to avoid extracting the background area as an object whose gray values have varied due to environmental conditions. The system proposed in the present paper extracts an object accurately using background difference, inter-frame difference in an image (time difference) and distribution of pixel values of an object area. Based on the information thus obtained, the system detects obstructions (stopped vehicles, objects dropped from running vehicles, etc.), tracks vehicles (lane changing, speed reduction, etc.), and grasps the traffic status on roads. This procedure allows real time processing on a personal computer, incurring low calculation cost. (A*)
Samenvatting