基于近眼红外图像的高精度瞳孔中心定位方法

High-precision Pupil Center Positioning Method Based on Near-eye Infrared Image

针对近眼红外相机采集的人眼图像存在反射光斑干扰和瞳孔部分遮挡导致瞳孔中心定位精度不高的问题,提出了一种高精度瞳孔中心定位方法。首先获取瞳孔部分感兴趣区域,减小后续计算量;再采用中值滤波去除椒盐噪声;接着利用阈值分割、形态学处理、获取最大连通区域和边缘检测去除背景区域及残留噪声,提取瞳孔边缘轮廓;然后采用光斑区域检测剔除光斑干扰点,判断瞳孔遮挡情况并去除非边缘点,最后利用改进椭圆拟合算法定位瞳孔中心。实验结果表明,该算法能够在瞳孔受反射光斑干扰和被部分遮挡情况下准确定位到瞳孔中心,总体精度达到98.2%。

Aiming at the problem of poor pupil center positioning accuracy due to the interference of reflected light spots and partial occlusion of the pupil in the human eye image collected by the near-eye infrared camera,a high-precision pupil center positioning method is proposed.Firstly,the area of interest in the pupil part is obtained to reduce the amount of subsequent calculation and the median filter is applied to remove the salt and pepper noise.Then,the tasks of eliminating the background area and residual noise and extracting the contour of the pupil edge are completed by performing threshold segmentation,morphological processing,acquiring the maximum connected area,and edge detection in sequence.After that,spot area detection is employed to eliminate spot interference points,and wipe off non-edge points according to pupil occlusion.Finally,an improved ellipse fitting algorithm is designed to locate the pupil center.Experimental results show that the algorithm can accurately locate the center of the pupil when the pupil is disturbed by the reflected light spot or partially occluded,with an overall accuracy of 98.2%.