The authors propose a new image processing method, consisting of the temporal differencing method, with certain modifications, for detecting moving objects. Generally, the temporal differencing method is subject to a problem in that distant objects cannot be detected because their movements in images are too small. To solve this problem, the authors used the appropriate interval of input images. The appropriate interval was obtained by applying the restrictive condition that objects move on a plane which is fixed relative to the camera. They applied the proposed method to their traffic surveillance system and it worked in real time. The experimental results are presented, and it is shown that the proposed method detects the vehicles for a 33% greater average tracking duration in the day image than does the conventional method.
Abstract