摘要
在基于视频处理的室外场景运动物体检测系统中,对运动物体的阴影进行检测与去除是一个关键环节。文中提出了一种基于纹理自相关和整数小波变换相结合的运动物体阴影检测与去除算法,算法中首先使用纹理自相关对运动物体的阴影进行预提取,再对阴影预提取结果进行统计判别,对判别为阴影误检的区域再进行基于整数小波变换的阴影再检测,最后将两次检测结果相结合实现对运动物体阴影的检测与去除。实验结果表明:文中的方法不仅能够准确地检测出与背景灰度差别比较大的运动物体的阴影,而且能够较好地检测出与背景灰度相近的运动物体的阴影,较好地克服了使用单一方法进行阴影检测与去除时常见的阴影误检问题,获得了很好的阴影检测与去除效果。
In outdoor moving object detection system based on video processing,the shadow of the moving object detection and removal is a key link.This paper proposes that a moving object＇s shadow detection and removal algorithm is based on texture autocorrelation and integer wavelet transformation.The algorithm firstly uses the texture autocorrelation to the pre-extraction of the shadow of the moving objects;secondly by use of the statistical discrimination methods it evaluates the shadow pre-extraction results;then by use of the integer wavelet transformation it re-extract the misjudgment of the shadow area;and finally by combining the two shadow extraction results it actualizes the shadow detection and removal of the moving object.Experimental results show that： the method not only can accurately detect the shadow of the moving object which is a large differences in gray in comparison with the background,but also can better detect the shadow of the moving object which is similar to the background in gray,so the method solves the common false detection question of the shadow when using the single method to detect the shadow and remove it,and get a wonderful shadow detection and removal results.
出处
《东莞理工学院学报》
2011年第3期46-52,共7页
Journal of Dongguan University of Technology
基金
广东省工业科技攻关计划项目"基于视频检测的关联交叉口群智能控制系统的研究与实现"(2005A10101001)
关键词
阴影检测
纹理自相关
整数小波变换
运动物体检测
shadow detection
texture autocorrelation
integer wavelet
moving object detection
