一种基于LBP和GLCM的单帧人脸近红外图活体检测方法  

Single Frame Face Liveness Detection from Near Infrared Image Based on LBP and GLCM

作　　者：张自友[1,2] ZHANG Ziyou(School of Physics and Electronic Engineering,Leshan Sichuan 614000,China;National Key Laboratory of Fundamental Science on Synthetic Vision of Sichuan University,Chengdu Sichuan 610064,China)

机构地区：[1]乐山师范学院物理与电子工程学院,四川乐山614000 [2]视觉合成图形图像技术国防重点学科实验室(四川大学),四川成都610064

出　　处：《乐山师范学院学报》2018年第12期5-12,73,共9页Journal of Leshan Normal University

摘　　要：文章提出了一种简单、快速、有效的人脸活体检测方法:通过对人脸近红外(NIR)图计算局部二值模式(LBP)图以增强光照鲁棒性,然后计算灰度共生矩阵(GLCM)及特征值实现数据降维;通过支持向量机(SVM)对降维后正负样本部分数据训练后得到分类模型,从而成功实现了基于单帧人脸近红外图的活体检测。文章的算法有效抑制了环境光照、图像明暗、人脸姿态变化等的影响,在开放数据集CASIA NIR Database以及自行采集的数据集上进行的实验取得了很好的效果。This paper presents a simple,fast and effective method of face liveness detection.The Local Binary Pattern(LBP)map is calculated by the Near Infrared(NIR)image to enhance the robustness of illumination.Then the eigenvalue of Gray Level Co-occurrence Matrix(GLCM)are calculated to reduce the dimension of the data.After training the positive and negative sample data,the classification model is obtained by Support Vector Machine(SVM).In this way,the single frame face liveness detection based on the near infrared image is successfully realized.The algorithm can effectively suppress the influence of illumination,image shading,and face pose changes and so on.The experiments on the open data set CASIA NIR Database and the self-collected data set have achieved good results.

关 键 词：人脸活体检测 照片攻击 近红外图像 DLIB LBP GLCM SVM 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]