一种基于模糊分类的模板匹配眼睛定位方法  

A Faster and More Accurate Template Matching Eye Location Method Based On Fuzzy Classification

作　　者：史慧荣[1] 张学帅[1] 梁彦[1] 张洪才[1] 程咏梅[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《西北工业大学学报》2005年第1期55-59,共5页Journal of Northwestern Polytechnical University

基　　金：国家自然科学基金 (6 0 372 0 85 );陕西省科学技术研究与发展项目 (2 0 0 3K0 6 - G15 );西北工业大学青年教师创新基金;西北工业大学研究生创业种子基金 (Z2 0 0 4 0 0 4 3)资助

摘　　要：针对眼睛定位问题 ,提出了一种基于模糊分类的模板匹配方法。在用一种合成的眼睛模板对人脸图像进行匹配得到眼睛定位点的备选集合的基础上 ,通过模糊分类技术引入双眼相对位置关系等知识信息 ,从备选集合中找出相似度最高的一组作为双眼的定位点。实验证明与传统模板匹配方法相比 ,该方法能够同时定位双眼 ,并能够显著提高定位准确率和定位速度 ,对戴眼镜、头部变化。To satisfy the need of real-time face-recognition system, eye location method must become faster and more accurate. This paper presents an eye location method that is one step nearer to meeting this need. In this paper, a fuzzy-classification-based template matching method is proposed for the eye location problem. We explain our proposed method in much detail in the full paper. Here we give only a briefing of our explanation. First, we synthesize a left and a right eye template as a synthetic eye template to improve the robustness to eye variety and reduce computation burden. A set of eye-points is obtained through matching synthetic eye template with face image. Then fuzzy classification is used to divide the similar eye points into two groups using the location of one eye relative to the other as constraining relationship. Finally, the two eye points are located by finding out the most similar pair. The experimental results show that the new method can locate the two eyes simultaneously, get higher precision and faster speed compared with the traditional template matching algorithm. Besides, even when there occur special conditions——such as glass-wearing face, pose-varied face, etc——which cause eye location by traditional algorithm not quite satisfactory, our proposed fuzzy-classification-based template matching method still has the robustness to give satisfactory location.

关 键 词：人脸识别 眼睛定位 模板匹配 模糊分类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]