Abstract: A new approach is presented to locate the position of pupil automatically. It sets the eyes searching area via the distribution of eyes in the face. Through Adaboost algorithm, the cursory position of human eyes can be got by searching the region-of-interest. Using image binarization and Canny algorithm, it can extract the contour of image. It can obtain the center coordinates of human eyes by ellipse fitting, which is the position of pupil. The experimental results demonstrate that the proposed method can locate pupil in different illumination and gesture with high rate.

Key words: image binarization, eye detect, ellipse fitting, pupil localization

摘要： 提出一种自动定位瞳孔的方法。该方法利用人眼在人脸的位置分布设置人眼的搜索区域，通过Adaboost（The Adaptive Boosting Algorithm）算法搜索先前设置好的区域，可以得到人眼所在粗略位置，利用图像二值化和Canny算法提取图像轮廓，通过对提取的轮廓进行椭圆拟合即可得到人的眼球中心坐标，即瞳孔的位置。实验结果表明该方法能准确定位瞳孔，对光照、姿态具有一定的鲁棒性。

关键词: 图像二值化, 人眼检测, 椭圆拟合, 瞳孔定位