一种鲁棒的人眼光斑定位算法  

A robust algorithm for human eye glint localization

作　　者：邓宏平[1] 李斌[1] 姚鹏[1] 张文聪[1] 庄镇泉[1] 

机构地区：[1]中国科学技术大学电子科学与技术系智能信息处理实验室,安徽合肥230027

出　　处：《中国科学技术大学学报》2011年第4期341-346,共6页JUSTC

摘　　要：光斑的检测和定位是视线跟踪系统中的一个重要环节,其定位精度直接影响视线跟踪系统的精度.目前常用的基于灰度二值化的光斑检测方法,容易受到光斑亮度变化和镜面反光等因素的影响.径向对称变换(RST)是一种有效的圆形目标检测技术,对上述干扰具有较好的鲁棒性,但计算代价偏高.为此提出了一种改进的径向对称变换(IRST),有效降低了算法的计算代价.在引入IRST实现光斑检测之后,再结合自适应二值化和重心法实现光斑的精确定位.实验结果表明,IRST算法的光斑检测鲁棒性明显高于基于灰度二值化的方法,而计算代价明显低于基于经典RST的光斑检测算法,最终的定位精度也优于其他同类算法.Glint detection and localization are important steps of gaze tracking and the accuracy of localization directly affects the precision of eye tracking system.At present,most approaches to glint detection are based on image binaryzation,which is sensitive to illumination changes and glasses reflection.Radial symmetry transform,which is robust to the disturbance above,is an effective circle detection method,but computationally expensive.An improved radial symmetry transform was proposed,which remarkably reduced the cost of computation.By incorporating the IRST of detecting the glint,the glint center was calculated using image adaptive binary and the gravity of glint.Experimental results show that the detection of glint is more robust than other methods based on image binaryzation,the cost of computation is lower than RST,and the localization accuracy is also better than other methods.

关 键 词：光斑检测 光斑定位 视线跟踪 径向对称变换 

分 类 号：R318[医药卫生—生物医学工程]