复杂背景下的人脸检测与跟踪系统  

A HUMAN FACE DETECTION AND TRACKING SYSTEM FOR COMPLEX BACKGROUNDS

作　　者：刘明宝[1] 高文[1] 

机构地区：[1]哈尔滨工业大学计算机科学与工程系

出　　处：《计算机研究与发展》1997年第S1期61-65,共5页Journal of Computer Research and Development

基　　金："863"计划

摘　　要：在复杂背景下进行人脸检测与跟踪是多功能感知的重要交互通道．文中设计并实现了一种基于特征子脸（ｅｉｇｅｎ－ｓｕｂｆａｃｅ）的人脸检测与跟踪系统．首先利用模板匹配的方法进行粗检测，设计了一种基于灰度分布的人脸模板，并在此基础上收集有效的反例样本集（非人脸样本集）．提出了特征子脸的概念，设计了有效的人脸器官空间基底，并给出了详细的求解方法及检测与跟踪策略．测试结果表明，本系统可以在任意背景下进行令人满意的人脸检测与跟踪．Detecting and tracking human faces in a complex background is one of the most important multimodal human machine interfaces. This paper presents a human face detection and tracking system which is based on eigen subfaces. A standard human face template is established and a method of template matching is used as the first step. The concept of eigen subface is then proposed and a set of effective bases for human face's organ space are designed. The eigen subface's resolution measure and the strategy of human face detection and tracking are also presented are designed. Experimental results show that this system could detect and track human faces satisfactorily in various unconstrained backgrounds.

关 键 词：人脸检测与跟踪 模板匹配 特征子脸 人脸器官空间 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]