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

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.