结合HSV空间阴影消除的高斯混合模型背景减除方法

Background subtraction method of Gaussian mixture model combined with HSV space shadow elimination

针对高斯混合模型背景减除后运动目标存在阴影的问题,提出了初始化参数快速估计方法避免了高斯混合模型参数估计易产生局部极大值的问题,在分析高斯混合模型背景减除结果中阴影区域亮度和色度变化特性的基础上,提出了将高斯混合模型背景减除结果映射到HSV空间并通过亮度和色度阈值消除阴影的H SV空间阴影消除方法。基于上述方法对不同光照强度下序列图像进行背景减除,实验结果表明本文方法能较准确地检测并消除阴影区域,平均目标检出率高于80%。

In order to deal with the shadow of the background subtraction results, we propose the fast estimation method of initial parameter, which avoids the problem that the Gaussian mixture model parameter would generate local maximum. In this paper, we mapped the result of Gaussian mixture model background subtraction to the HSV color space, and a method of shadow elimination is proposed based on the analysis of the brightness and chromaticity variation of shadow areas. Experiments under different light intensities show that this method can eliminate the shadow area more accurately, and the detection rate of the target is generally higher than 80%.