改进的遮挡条件下瞳孔检测方法

受到眼睑、睫毛、反光等干扰,提取的瞳孔往往无法近似成标准的椭圆,给瞳孔精确检测带来了难度。针对遮挡问题,提出了一种改进的遮挡瞳孔精确检测方法,该方法首先获取瞳孔图像的感兴趣区域,在对该区域进行阈值分割的基础上,通过滤波去噪和扫描线法去除光源反光的影响;根据沿遮挡瞳孔的分界线进行水平旋转,应用内接平行四边形法确定瞳孔的中心;根据圆形检测理论,提出一种中心修正算法,将求得的两个圆的交点作为眼睑与瞳孔的交点,计算出旋转角度,从而精化瞳孔中心的定位;最后利用五点法计算得到的瞳孔椭圆参数作为初始值,通过最小化拟合椭圆与边缘点之间的欧氏距离,使得该椭圆在非线性最小二乘意义下是最优的。实验结果表明,该方法具有良好的鲁棒性和准确性,在遮挡不超过瞳孔面积一半的情况下,能得到较为精确的瞳孔位置和边界。

Improved method for the pupil measurement under occlusion

Pupil can not be extracted as a similar standard ellipse because of some interferences caused by reflections, eyelashes and eyelids, which have some difficulties to measuring pupil accurately. In order to solve the problem of eyelid occlusion, an improved method of pupil measurement under eyelid occlusion is presented. It acquires the region of interesting on pupil image and removes reflections by filtering and Sweepline based on threshold segmentation. Then, the center of the pupil is determined using inscribed parallelogram after rotating according to the dividing line for the eyelid. A proposed algorithm of center correction based on circle detection can detect the intersections of two circles, which can be used as the intersections between the pupil and eyelid in order to compute the angle of the pupil and refine the center. Furthermore, the initial parameters of the pupil ellipse obtain from the five-point method and the ellipse is optimal in the sense of non-linear least squares by minimizing the Euclidean distance between fitting ellipse and edge points. Experimental results demonstrate, this method achieves good performance in terms of robustness and accuracy and obtains accurate locations and boundaries in the case of occlusion no more than half of the pupil area.