结合RGB颜色特征和纹理特征的消影算法

Shadow Elimination Algorithm of Combination of RGB Color Feature and Texture Feature

在视频监控的运动检测中,运动阴影常被误认为视频对象。为了解决这个问题,提出了一种结合RGB颜色特征和纹理特征的运动阴影检测方法。利用背景差和多帧差结合的方法,获得包含运动阴影的前景分割,在此基础上利用阴影区域中RGB特征的相似性(其RGB向量夹角较小)与局部二值纹理模式(Local Binary Pattern,LBP)相结合,进行阴影检测与去除;使用形态学的方法去除像素突变的边缘部分。实验表明,该方法在运动目标检测及跟踪方面取得了较好的效果,与其他消影算法比起来,速度较快。

In video motion detection,moving shadow is often mistaken as video object. In order to solve this problem,the shadow detec-tion method of combining RGB color feature and texture feature is proposed. Video target segmentation is achieved by background differ-ence and multi-frame difference. The RGB similarity （ RGB vector angle becomes smaller） in shadow area and the Local Binary Pattern （ LBP） is employed to detect and eliminate the shadow. The morphological method is used to eliminate the edges of pixel mutation. The experimental results demonstrate the proposed method has good effect on target detection and tracking,is faster than others.