单摄像机视线跟踪  

Eye-gaze tracking with one camera

作　　者：刘瑞安[1] 靳世久[1] 宋维[2] 张希坤[2] 吴晓荣 

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072 [2]天津师范大学物理与电子信息学院,天津300074

出　　处：《计算机应用》2006年第9期2101-2104,共4页journal of Computer Applications

基　　金：天津市高等学校科技发展基金资助项目(20051017)

摘　　要：提出一种能对环境光强变化、用户头部位置变化自动适应的视线跟踪方法,减少了视线跟踪系统对使用者的头部限制。改进了视线跟踪算法:依据直方图,给出了一种自适应阈值提取方法;依据瞳孔边界点的灰度信息、梯度信息及瞳孔边界曲线的平滑信息综合判据,给出了一种提取瞳孔边界点的一维算法;给出了一种随机化椭圆拟合算法;讨论了去除眼皮、眼睫毛及光斑干扰的方法。实验结果验证了算法的有效性。An eye-gaze tracking method, which can make this system be automatically adapted to external lighting changes and the user＇s head movements freely, was put forward to alleviate the restriction on the user＇s head movements. In this paper, we concentrated on development of the algorithms. Based on the brightness histogram of the eye image, the adaptive thresholds for pupil extraction and corneal reflection were extracted. On the basis of the brightness, gradient of the edge points of the pupil and the smoothness of the pupil boundary, the edge points could be obtained by a 1D line search. We applied a new randomized method of ellipse fitting to the edge points. A solution to cope with eyelashes, eyelids and glints in the pupil area was also discussed. Experiments show that the improved algorithm is effective.

关 键 词：视线跟踪 角膜反射 自适应阈值 直方图 椭圆拟合 校准 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]