Abstract: A problem has occured when the video analysis work is done because of the big data of the surveillance video, it is convenient to analyze the video if the positive face has found. Discriminant of the positive face uses the method based on integral channel features and Adaboost, in the training section, the Adaboost uses the extracted integral channel features to train the classifier, which is used to the rapid discriminant of the positive face of the face detection window. Experimental results show that this method can meet most surveillance video with good computing speed and accuracy.

Key words: surveillance video, video analysis, integral channel features, Adaboost

摘要： 视频监控领域由于其数据量巨大使得对视频分析变得更加困难，找到视频中正面人脸图片能够使得分析变得简单。正面人脸判别采用了基于积分通道特征和Adaboost的方法，通过提取积分通道特征，利用Adaboost训练分类器，对人脸检测窗口进行正面人脸快速判别。实验结果表明，该方法具有比较好的判别速度和精度，能够应用于大部分监控视频。

关键词: 监控视频, 视频分析, 积分通道特征, Adaboost