一种用于虹膜识别的眼睑和睫毛检测算法  

Eyelid and Eyelash Detection Algorithm for Iris Recognition

作　　者：周俊[1] 罗挺[2] 李敏[1] 余丛斌[1] 

机构地区：[1]后勤工程学院信息工程系,重庆401331 [2]后勤工程学院训练部模拟中心,重庆401331

出　　处：《武汉理工大学学报（信息与管理工程版）》2011年第2期175-179,共5页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家自然科学基金资助项目(70801062)

摘　　要：为减少眼睑和睫毛对虹膜识别率的影响,提出了一种用于虹膜识别的眼睑和睫毛检测算法。应用灰度形态学运算去除睫毛、光斑对眼睑区域的干扰,提取上眼睑区域每列灰度最小值点作为边缘点,对下眼睑区域进行梯度变换,根据最大梯度积分区间确定下眼睑边缘点;对眼睑边缘点进行最小二乘抛物线拟合上、下眼睑;利用眼睑抛物线在邻域内上下移动,当灰度突变时精确定位上下眼睑。根据上眼睑抛物线下移后抛物线像素点灰度的分布,自适应生成阈值对睫毛进行分割。实验结果表明:该眼睑算法能有效、快速地检测眼睑,与Hough变换算法相比,准确率相当,速度提高约4 s;睫毛检测算法能有效、简便和最大限度地检测出睫毛。In order to reduce the influence of eyelid and eyelash on iris recognition rate,eyelid and eyelash detection algorithm for iris recognition was proposed.Firstly,gray-scale morphological operation was applied to remove interference of eyelash and eyelid.A point which has minimum grayscale in each column was extracted from upper eyelid area as an edge point.Gradient transform was conducted on lower eyelid;the edge points were determined according to the integral interval which has maximum gradient.Then eyelid edge points were used to fit eyelids with the least-square parabola method.Finally,moving the eyelid parabolic in the neighborhood,the upper eyelid and lower eyelid were accurately positioned by gray mutations.Moving down the upper eyelid parabolic,gray distribution of pixels in eyelid parabolic was self-adapted to create threshold,by which eyelash was segmented.Experiment results show that this algorithm can effectively and quickly detect eyelid.Compared with the Hough transform algorithm,the proposed algorithm has the same accuracy,but speed can be accelerated by more than 4 s.Eyelash detection algorithm can effectively and conveniently detect eyelash in greatest degree.

关 键 词：虹膜识别 眼睑检测 最小二乘拟合 睫毛检测 自适应阈值 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]