基于图像梯度和改进椭圆拟合算法的视线追踪方法  

Line of sight tracking method based on image gradient and improved ellipse fitting algorithm

作　　者：王鹏 苑硕 董鑫 陶文杰 WANG Peng;YUAN Shuo;DONG Xin;TAO Wenjie(School of Electrical and Information Engineering,Changzhou Institute of Technology,Changzhou 213032,China;Innovation and Development Center of Qingdao,Harbin Engineering University,Qingdao 266000,China;School of Electrical and Electronic Engineering,Harbin University of Science and Technology,Harbin 150080,China)

机构地区：[1]常州工学院电气信息工程学院,江苏常州213032 [2]哈尔滨工程大学青岛创新发展基地,山东青岛266000 [3]哈尔滨理工大学电气与电子工程学院,哈尔滨150080

出　　处：《扬州大学学报（自然科学版）》2023年第4期48-53,60,共7页Journal of Yangzhou University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61602063).

摘　　要：针对单摄像机视线追踪系统检测精度低且速度慢等问题,提出一种基于图像梯度和改进的椭圆拟合算法的视线追踪方法.首先,利用改进的人脸识别算法和“三庭五眼”的人脸比例算法对预处理后的图像信息进行人脸和眼睛部位的检测定位;其次,采用图像梯度算法定位人眼的瞳孔中心点,再分别利用改进的星射线算法与椭圆拟合方法确定虹膜边缘和虹膜中心,通过中心坐标评判完成瞳孔中心定位;最后,使用坐标法判定视线方向.实验结果表明,该方法可有效降低头动、眼睑和眼镜等因素造成的检测误差,并且能够准确判断人眼视线方向,具有较好的鲁棒性和实时性.Aiming at the low detection accuracy and slow speed of single-camera line of sight tracking system,a gaze tracking method based on image gradients and improved ellipse fitting algorithm is proposed.The method involves face and eye detection using an improved face recognition algorithm and the“three vertical,five horizontal”face proportion algorithm.Eye center localization is achieved using the image gradient algorithm.The iris edge and center are determined using an improved starburst algorithm and ellipse fitting method.Gaze direction is determined using a coordinate-based approach.Experimental results demonstrate that this method can effectively reduce the detection errors caused by head movements,eyelids,glassses and other factors.It can determine gaze direction with good robustness and real-time performance.

关 键 词：视线追踪 瞳孔中心定位 图像梯度 椭圆拟合 人脸检测 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]