一种近红外图像中人眼定位的方法  

A method for eye localization in infrared images

作　　者：闫敬文[1] 张田[1] 赵慧民[2] YAN Jingwen ZHANG Tian ZHAO Huimin(College of Engineering, Shantou University, Shantou Guangdong 515063, China Department of Electronics and Information, Guangdong Polytechnic Normal University, Guangzhou Guangdong 510665, China)

机构地区：[1]汕头大学工学院,广东汕头515063 [2]广东技术师范学院电子与信息学院,广东广州510665

出　　处：《中国体视学与图像分析》2016年第4期381-385,共5页Chinese Journal of Stereology and Image Analysis

基　　金：广东省自然科学基金重点项目(2016A030311013);广东省教育厅省级重大资助项目(2014KZDXM060);广东省普通高校国际合作重大项目(2015KGJHZ021)

摘　　要：针对在可见光中人脸检测受光照影响的问题,提出基于Adaboost的近红外光人脸检测与人眼定位算法。首先使用基于Haar特征的Adaboost算法得到人脸区域,通过统计大量的人眼相对人脸位置的分布确定人眼待测区域。为了减少亮瞳与非亮瞳以及左右眼的差异影响,使用亮瞳、非亮瞳、左眼、右眼检测器,然后再用Haar特征的Adaboost算法在人眼待测区域进行人眼定位。实验结果表明,该方法的准确率高,速度快,达到实时性的要求。In view of the problem of face detection in visible light, a novel method for face detection and eye localization algorithm in infrared images based on Adaboost was p tion based on Haar feature ured is determined by the and Adaboost distribution of algorithm is used for locating ropo the sed in this paper. Face detec- face area. The eye to be meas- large number of human eyes relative to the face. In order to dealwith the difference between the bright pupil and the non bright pupil and the left and right eyes, detector separate into four parts, bright pupil eye detector, thenon bright pupil eye detector, left and right eye de- rector. Experimental results show that this method is capable of achieving high accuracy, fast and real - time performance.

关 键 词：ADABOOST算法 HAAR特征 人脸检测 人眼定位 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]