结合直方图反投影的多特征运动阴影检测算法

针对视频监控中运动阴影影响目标检测跟踪准确性的问题,提出一种结合直方图反投影的多特征运动阴影检测算法。首先利用背景减除法,得到前景区域并进行初步筛选;然后在背景区域建立亮度、颜色、梯度特征的联合直方图,以反投影的方式投影到前景区域得到运动阴影概率图;最后结合空间一致性和滞后阈值,对概率图进行分割得到运动阴影区域。与典型算法进行对比的实验结果表明,本文算法能够有效区分阴影与目标,适用于实时的运动目标检测与跟踪。

Multi-feature histogram back-projection for moving shadow detection

As the accuracy of target detection an d tracking is disturbed by moving shadow in video surveillance,a moving shadow detection algorithm combined with histogram back-projection and multi-feature is proposed in this paper.At first,Foreground areas are extracted by applyin g background subtraction,and candidate shadow regions are selected immediately.Then,the join t histogram including the intensity,color and gradient of background region according to the candidat e shadow area is established.Shadow probability image was obtained by back-project the joint hi stogram to the foreground area.Finally,the moving shadow region is detected by segmenting the probability image with the spatial consistent and hysteresis threshold.Experimental results show that the proposed algorithm is applicable to real-time moving object detection and can tracking and separate the object and shad ow efficiently.