Abstract: It is very necessary to apply motion detection to video surveillance in order to enhance the performance of system，and human motion detection is the foundation of advanced operations.An enhanced algorithm in which background image estimation based self-adaptation is combined with current consecutive frames subtraction，which can detect motive region rapidly and exactly.Afterward，it can realize human motion detection quickly and reliably by face detection which is based on the consecution of time domain in video images and skin information of face.According to real-time quality，presented the veracity and reliability of video surveillance is presented in this paper.The result of experiment demonstrates that the algorithm is robust in complex situation and adapts to video surveillance.

Key words: video surveillance, motion detection, two consecutive frames subtraction, background subtraction, skin model

摘要： 为了提高视频监控的实时性、准确性和可靠性，引入运动目标检测非常必要，而在此基础上的人运动检测更是后续各种高级处理的基础。根据视频监控的特点，采用一种基于自适应背景图像估计与当前多帧图像的混合差的算法来实现快速精确地检测和提取运动目标区域，并充分利用视频图像的时域连续特性和人脸肤色信息，实现快速可靠的人脸定位，从而准确定位人运动区域。实验表明，该算法对人的运动检测在光线、姿势变化等情况下具有良好的鲁棒性，适于实时监控系统的应用。

关键词: 视频监控, 运动检测, 帧间差分, 背景差分, 肤色模型