摘要
针对面部认证系统中活体检测正确率要求高的应用需求,设计出一套适于特定环境的面部认证系统,即先进行面部身份识别,再对识别后的面部图像进行活体检测,为特定个体建立特定的检测模型,具有很强的特指性.针对该特定系统中合法用户的面部图像,采用局部二值模式（LBP）、方向梯度直方图（HOG）、灰度共生矩阵（GLCM）和哈尔小波变换（HAAR）等方法进行特征提取,通过对不同个体建立不同支持向量机网络进行活体验证.在公开的NUAA数据库上进行验证,正确检测率可达100%,实验结果表明：基于特定环境面部认证系统的活体检测模块不仅检测准确率高,而且能够满足特殊定制应用环境的需求.
To satisfy the high accuracy demand of liveness detection,a face authentication system（FAS） for specific environment was designed.First the face was identified,and then a liveness detection was implemented on the identified image.A specific test model was established for each specific individual with a strong pertinence.Local binary pattern（LBP）,histogram of oriented gradient（HOG）,gray-level co-occurrence matrix（GLCM）,Haar-like（HAAR） methods were used to extract features for each legal user＇s liveness facial images and photo facial images,and then a support vector machine（SVM） net was used for the liveness proof.The proposed method was validated on a public database of NUAA with the correct detection rate of 100%.The experimental results show that the detection module based on special environmental not only has a high detection accuracy,but also can meet the needs of special environment.
作者
李素梅
秦龙斌
胡佳洁
Li Sumei;Qin Longbin;Hu Jiajie(School of Electrical Automation and Information Engineering,Tianjin University,Tianjin 300072,China;Changdu Public Security Bureau,Changdu 854000,Xizang China)
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2018年第11期87-91,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国际(地区)合作与交流资助项目(61520106002)
国家自然科学基金资助项目(61471262)
关键词
活体检测
面部认证系统
特定环境
支持向量机
特征提取
liveness detection
face authentication system
specific environment
support vector machine
feature extraction
