Abstract: Eye movement is an important approach to get and promulgate the information of brain.In recent years，most researches are to model eyeball as a circle or an ellipse，to use Hough transform detect it，for locating the pupil center.For the traditional Hough transform（including random Hough transform） has a poor fristgerecht，in order to get more accurated eyemove information，the rate of frames gathering usually has to be over 100 f/s.So，by the special features of dark pupils and approximate circle of pupil，this thesis uses Video Oculographic（VOC） to gather the images，improves the traditional Hough transform，extracts the edge of pupil binary image，chooses three points on the boundary to get a parameter set，and replaces the traditional voting method with the mathematic expectation，calculates the mathematic expectation to get the center of pupil.The experiment proves that the method is feasible，accurate and stable.

Key words: circle detecting, pupil detection, center locating, Hough transform, dark pupil

摘要： 眼球运动是获得和揭示大脑工作信息的重要途径。近年来的研究多将眼球建模成圆或椭圆，利用Hough变换检测圆或椭圆，从而达到瞳孔中心定位的目的。传统的Hough变换（含随机Hough变换）的实时性较差，而为了获得更加准确有效的眼球运动信息，采集帧率往往需要超过100 f/s。为此，利用VOG的暗瞳特性及二值化后瞳孔图像圆的唯一性，通过Video Oculographic（VOG）设备采集眼部视频，对Hough变换进行改进，对预处理过的二值图像进行边缘提取，同时选取边缘上的三点通过点Hough变换得到一个参数组，再用数学期望替代传统Hough变换投票方式，对所有参数组求数学期望，得到最终瞳孔中心。实验证明，该方法可行、精确、稳定。

关键词: 圆检测, 瞳孔检测, 中心定位, 霍夫变换, 暗瞳