摘要： 在雨天交通视频中，车辆同时受阴影和水汽拖尾的干扰。为此，在分析目标车辆、路面、阴影和水汽拖尾区域特性的基础上，提出一种融合颜色相关性和纹理差异的阴影检测方法。由当前点与背景点的HIS向量点积测量颜色相关程度，检测出车辆目标和水汽拖尾区域，利用当前区域和背景区域的纹理差异区分深色车辆和阴影。实验结果表明，该方法能较好地检测雨天交通视频中运动车辆的投射阴影，且能有效去除水汽拖尾，从而保证车辆的正确分割。

关键词: 自动事件检测, 阴影去除, 颜色相关性, 纹理差异

Abstract: In rainy highway environment, segmentation of vehicle is disturbed by both cast shadow and the water tailing region. For solving this problem, on the analysis of the characteristics of target vehicle, road, shadow and water vapor trailing area, this paper presents a shadow detection method based on color correlation and texture difference. Color correlation is estimated between current pixel and background pixel through dot product of Hue, Saturation, Intensity(HIS) vectors to detect vehicle object and water tailing region. The texture difference based on gradient of current region and background region is calculated to distinguish dark vehicle from cast shadow. Experimental results show that this method can effectively eliminate both the cast shadow and the water tailing region of moving vehicles in rainy highway video, it can ensure the accurate segmentation of vehicle objects.

Key words: Automatic Incident Detection(AID), shadow elimination, color correlation, texture difference

中图分类号: 

• TP393