TRAFFIC MEASUREMENT WITH A ROADSIDE VISION SYSTEM - INDIVIDUAL TRACKING OF OVERLAPPED VEHICLES

Traffic measurement is an essential technique in moving toward the next stage of vehicular transportation, one which will be more advanced in terms of both safety and efficiency. For this technique, image sensors are more advantageous than conventional ones due to their portability and the wide variety of information they provide. But up to now, however, vehicles overlapping each other within the image plane have caused problems with measurement accuracy in such sensors. The authors, therefore, developed a traffic measurement algorithm which features individual vehicle tracking using a rectangular model. They operated the algorithm in real-time on a Realtime Vision System (RVS) to utilize traffic volume to evaluate the algorithm performance.