A Novel Traffic Flow Detection Method Using Multiple Statistical Parameters

This paper proposes a novel method of traffic flow detection based on multiple statistical parameters. This method processes only the image of the observation window, which is preset on each land of the road. The gray value variance of current image and constructed background is employed as the main statistical parameter to detect the vehicle. The texture-based statistical parameter is applied as a supplement to enhance the robust against the moving shadow. The results of surveillance video test demonstrate that this method is efficient, accurate and robust against moving shadow.