基于纹理不变性的车辆阴影处理方法(英文)

A New Vehicle Shadow Handler Based on Texture Invariance

由于视频蕴含丰富的信息,视频监控与检测技术在ITS中应用得越来越广泛.视频测量交通参数,需要从序列图像分割出车辆.车辆分割面临的一个难点就是检测与车辆连在一起的阴影.由于阴影在很多特性上与车辆一致,阴影经常被当成车辆一部分,影响车辆分割的稳定性与准确性.一般地,路面与车辆在纹理结构上存在较大的差异.文中推导了纹理在光照变化情况下的一种不变特性——极点分布,并提出了一个基于极点分布不变性的车辆阴影检测算法.这种算法能够精确地检测视频图像车辆的阴影,检测出因各种原因被误为车辆的路面,检测出车灯照射产生的路面亮斑,从而为车辆的精确分割和交通参数测量创造有利条件.

Video-based surveillance and measuring have been employed more and more widely in ITS （Intelligent Transportation System） because of the rich information content contained in video. The vehicles need to be segmented from the video images to measure traffic parameters. Because of the similarities between the vehicles and their shadow ,shadow is often segmented as vehicles. This paper derived an invariance property of texture-extremum distribution in changed illumination environment and presented a novel method for detecting shadow based on the extremum distribution invariance. This method can detect and remove the shadow of vehicles and the light spot caused by vehicle lamp or the other road surface accurately and can be performed fast and is insensitive to the illumination rarities.