Efficient background updating scheme for real-time traffic monitoring

Background updating is an important problem in dynamic scene analysis. It is critical to be able to distinguish between long-term gradual changes in the background and short-term rapid changes resulting from moving objects in the scene. In this paper we propose an efficient background updating scheme for real-time traffic monitoring. In particular, we address the sleeping person problem, which arises frequently in the context of real-time traffic monitoring. The proposed scheme combines two levels of reasoning: low-level reasoning based on pixel status analysis and high-level reasoning based on moving object correspondence analysis. The proposed scheme is robust and fast enough to satisfy the real-time constraints of traffic monitoring.