基于OEDLBP的人脸欺诈检测算法研究

Research on face spoofing detection algorithm based on OEDLBP

人脸识别技术提高了身份验证的效率,给用户带来快捷方便的体验.但是,当有人试图伪造用户人脸通过系统验证时,就会威胁合法用户的信息和财产安全.针对打印攻击和视频攻击,提出了一种基于奇偶数位像素差异的描述子OEDLBP,该算法将局部偶数位差值模式(EDLBP)、局部奇数位差值模式(ODLBP)和全局特征模式(GBP)结合,利用空间金字塔算法统计彩色图像通道内和通道间特征,将提取到的特征进行融合并用SVM对真假人脸进行分类,在CASIA-FASD、Replay-Attack和Replay-Mobile 3个人脸反欺骗数据库中取得了较好的实验效果.

Face recognition technology has improved authentication efficiency and given users a fast and convenient experience.However,when someone tries to fake the user′s face to pass the system verification,it will threaten the user′s information and property security.This paper proposes a descriptor based on odd and even bit difference local binary pattern(OEDLBP)for print and video attacks.The algorithm combines the even-bit difference local pattern(EDLBP),the odd-bit difference local pattern(ODLBP),and the global feature pattern(GBP),and uses the spatial pyramid algorithm to count the intra-channel and inter-channel features of the color image.Then,the extracted features are fused and then classified by SVM.This approach achieves good performance in three challenging face anti-spoofing databases:CASIA FASD,Replay-Attack,and Replay-Mobile.