中文摘要：

针对经典Vibe算法在运动目标检测时存在鬼影、阴影和噪声干扰的问题，提出了一种融合混合高斯模型的改进的Vibe算法。在背景初始化阶段，采用五帧差分算法与八邻域的像素值填充获得的真实背景并且消除鬼影现象；通过混合高斯模型权值与Vibe随机取样概率相结合进行背景更新，将得到的运动目标进行形态学处理，使运动目标更加清晰；最后，在YCbCr颜色空间进行阴影消除。实验结果表明，改进后的Vibe算法不仅能够有效地去除鬼影，并且在消除阴影与噪声方面取得了良好的效果。

英文摘要：

To solve the problems of the classical Vibe algorithm, such as ghost, shadow and noise interference in moving object detection, an improved Vibe algorithm which combines Gaussian mixture model was put for-ward. Firstly, by using five frame difference algorithm and eight-neighbor pixel value to fill in the background initialization phase, the true background is obtained and the ghosting phenomenon is eliminated. Then the background is updated by combining the Gaussian mixture model and Vibe random sampling probability, the moving targets are clearer with morphological method. And the shadow is eliminated in the YCbCr color space. The test results show that the improved Vibe algorithm not only can effectively eliminate the ghost image, but also has good effect in eliminating the shadow and noise.