面向实时交通视觉监控的综合动态背景更新方法

Comprehensive dynamic background updating method for real-time traffic visual surveillance

为了从复杂的交通场景中获取高质量的背景图像,提出了一种综合的动态背景更新方法。同时使用了帧差信息和提取的高层对象状态信息对混合高斯背景模型进行选择性更新,克服了因较长时间停车对背景的影响,并能及时消除由于背景物体移出造成的鬼影。对实时交通视频的处理效果表明,该方法兼具良好的自适应性与鲁棒性。

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.