Counting cars by tracking moving objects in the outdoor parking lot

At the parking lot of the expressway, the information concerning the state of congestion is offered to drivers. We propose the method of counting cars by using the tracking of moving objects from the sequential images taken at the outdoor parking lot. In this paper, we describe a loci extraction method by linkage and separation of the moving regions detected from time differential images, and a counting method by recognition of a car from the features of the extracted locus. We develop the system by combining a personal computer with a high speed image processor. Both the robustness and practicality of real time processing, at 0.2 second per frame, are confirmed through experiments using VTR real parking lot images.<>