一种基于监控视频的有效的人脸识别方法  

An Effective Face Recognition Method Based on Surveillance Video

作　　者：武琦 王夏黎[1] 王博学 赵晓娜 WU Qi;WANG Xia-li;WANG Bo-xue;ZHAO Xiao-na(School of Information Engineering,Chang’an University,Xi’an 710064,China)

机构地区：[1]长安大学信息工程学院,陕西西安710064

出　　处：《计算机技术与发展》2018年第12期59-61,66,共4页Computer Technology and Development

基　　金：国家自然科学基金(61473220);中国博士后科学基金(2012M521729)

摘　　要：现如今的视频监控技术在交通和安全领域已经得到了广泛的应用,其中人脸识别在视频监控中是一个重要的研究内容。与静态图像相比,基于视频图像序列的人脸识别具有更大的灵活性,由于监控视频中的人处于移动状态,通过摄像头截取得到的人脸图像可能存在模糊,分辨率较低等情况。为了提高对监控视频中低分辨率人脸图像信息处理的准确率,首先通过超分辨率迭代重构方法将低分辨率图像重构为高分辨率图像,然后利用Harr-Like特征和Adaboost算法构造一些弱分类器实现对人脸的检测,最后通过主成分分析法进行数据降维完成人脸识别。利用校园内实际监控视频进行实验,实验结果证明用超分辨率迭代重构后的人脸图像进行识别的准确率明显优于直接进行PCA的传统方法。Nowadays,video surveillance technology has been widely used in the field of traffic and safety,of which face recognition is an important research content.Compared with the static image,the face recognition based on the video image sequence has more flexibility.Because people in the monitoring video are in a moving state,the face image captured by the camera may exist fuzziness,low resolution and so on.In order to improve the accuracy of low-resolution face image information processing in surveillance video,we reconstruct low-resolution image into high-resolution image by super-resolution iterative reconstruction method firstly,then use the Harr-Like feature and Adaboost algorithm to construct some weak classifiers for face detection.Finally we carry on the data dimensionality reduction through principal component analysis for face recognition.The experiment on actual surveillance video in campus shows that the recognition accuracy of face images reconstructed by super-resolution iterative reconstruction is better than the traditional PCA method.

关 键 词：智能视频监控 超分辨率重构 人脸识别 HAAR特征 PCA算法 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]