基于自适应高斯概率的阴影检测

SHADOW DETECTION BASED ON ADAPTIVE GAUSSIAN PROBABILITY

提出一种基于自适应高斯概率的运动阴影检测方法．用高斯模型对阴影建模，并利用场景中的典型阴影区域初始化均值和方差．计算背景减除得到的前景点的阴影似度概率作为是否阴影的判据，被判为阴影点的像素将作为均值和方差的学习样本用来调整和更新参数以适应场景的动态变化．实验结果表明本文算法具有较高的阴影检测率和较低的误检率．

A scheme to detect moving cast shadows based on adaptive Gaussian probability was proposed. Gaussian model was used to model the distribution of shadow pixels, and typical shadow segments manually segmented were used to initiate the mean and the deviation. The pixels in the foreground were classified shadow points or object points according to their corresponding Gaussian probability values. The shadow pixels were then employed to update the Gaussian model in order to adapt to the scene changes proposed algorithm gives rather high shadow pixels detection rate and low false ~ Experimental results show that the alarm rate.