基于Taubin拟合方法的眼睛定位算法

The Method of Eye Location Based on Taubin Fitting

眼睛的定位及特征的提取在表情的理解、人机交互、身份识别、人脸检测等方面有重要的作用。直接用最小二乘法拟合椭圆的方法定位眼睛时,容易将不该生成椭圆解的地方拟合出了椭圆,Hough变换所需计算量大,占用内存大。针对上述问题,提出采用Taubin拟合与随机Hough变换(RHT)算法相结合方法获得眼睑的椭圆方程,最后把拟合椭圆的中心点作为眼睑中心点,拟合的椭圆作为眼睑的边缘。实验结果表明,该算法定位眼睛的平均检测率达94.6%,平均检测时间为0.12秒。

Eye location and feature extraction play an important role in the applications of expression comprehension, human - computer interaction, identification recognition and face detection etc. Due to the fact that di- rectly locating eye with least square method may fit ellipse in the place where ellipse should not fit, and Hough transform needs numerous calculations and takes up very large part of memory ; the new method of eye location, which obtains elliptic equations of eyelids on the basis of Taubin fitting and Randomized Hough Transform, is proposed. Finally , the center of the ellipse is regarded as the center of the eyelid , and the edge of fitting ellipse is regarded as the edge of the eyelid. Experiments show that the average testing rate is about 94.6%, and the average testing time is 0.12 seconds.