关键词: 背景模型, 实时交通视觉监控, 背景更新, 混合高斯分布, 鬼影

Abstract: To extract good background image from complex traffic scenes, a comprehensive dynamic background updating method was proposed. It used the information of both frame-difference and high-level object state to update the mixed Gaussian background model. It could avoid the effect of the prolonged traffic standstill on the background and eliminate the ghosts caused by the move-away of background objects. The results of experiment on the real-time traffic video show that the method is self-adaptive and robust.

Key words: background model, real-time traffic visual surveillance, background updating, mixed Gaussian distribution, ghost