Adaptive background estimation for real-time traffic monitoring

In this paper we propose an adaptive background estimation algorithm for outdoor video surveillance system. In order to enhance the adaptation to the slow illumination changes and variant input noise in long-term running, an improved Kalman filtering model based on local-region is discussed to dynamically estimate a background image, in which the parameters are predicted by a RLS adaptive filter accurately. The experiment results on real-world image sequences show that the algorithm performs robustly and effectively.