基于视频的人脸验证  

Video Based Face Verification

作　　者：庄莉[1] 艾海舟[1] 徐光祐[1] 

机构地区：[1]清华大学计算机科学与技术系

出　　处：《电子学报》2002年第8期1222-1225,共4页Acta Electronica Sinica

基　　金：8 63项目 (No.863 30 6 ZT0 3 0 1 1 )

摘　　要：本文提出了一种基于视频的人脸验证方法 .采用立体视觉方法初步将人脸区域与背景分割开 ,再根据多关联模板匹配方法精确定位人脸 .对定位后的人脸区域抽取特征器官位置 ,再依此裁剪出人脸样本 .从视频流中收集人脸样本 ,训练支持向量机 (SVM)作为验证器 .实验表明该方法在复杂的现场环境下是有效的、鲁棒的 .A video-based face verification method is proposed. Stereo vision is used to roughly segment face area from its back-ground, and then multiple-related template matching method is used to locate the face area finely. The face organs are extracted from the located face area and accordingly the face sample is cut out. Face samples are collected from video sequence, and a support vector machine classifier is trained on this sample set against other face samples as his verifier. Experimental results demonstrate the feasibility and robustness of this algorithm in complex real environment.

关 键 词：人脸检测 人脸特征抽取 支持向量机 SVM 模式识别 视频序列 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]