一种利用背景光流特征的虚假人脸检测方法

doi:10.19665/j.issn1001-2400.2021.05.011

摘要/Abstract

摘要：

人脸识别系统会受到照片视频的虚假人脸攻击,但手持照片、视频攻击存在抖动现象,以光流表征运动信息,提出一种利用背景光流特征的假人脸检测方法。该方法聚焦真实人脸背景与虚假人脸背景之间的运动差异性,分析人脸外区域的光流角度分布,背景对比区的运动状况,通过评估对比区的运动一致性虚假人脸。新方法在公开数据集ReplayAttack和CASIA-FASD中的测试准确率分别为97.87%和 90.95%,可有效甄别含背景区手持照片视频的虚假人脸攻击。

关键词: 虚假人脸检测, 背景运动, 光流计算, 直方图

Abstract:

The face recognition system is vulnerable to spoofing attacks by presenting photosor videos of a valid user,but there is a hand shaking existing when the illegal user holds the photosor videos.Focused on this phenomenon to represent motion information by optical flow features,an ovelface anti-spoofing method which exploits the background of optical flow features is proposed.The method took advantage of the background motion difference between the registered user and the invalid user to analyze the optical flow angle distribution of outside the face in multi comparing regions.And the similarity function is built to measure the correlation degree between the discograms of optical flow angle distribution in comparing regions.According to the measurements of the background comparing regions,their motion consistency have been evaluated,then the real face or the attacked face will be detected.Experiments have been done on the publicface anti-spoofing database of the Replay Attack and CASIA-FASD.The accuracy of the new method is 97.87% and 90.95%,respectively,which shows that the new method can effectively detect the attacker hold photosor videos with the background region.

Key words: face anti-spoofing, background motion, optical flow computation, histograms

中图分类号:

TP391

孔月萍,刘楚,朱旭东. 一种利用背景光流特征的虚假人脸检测方法[J]. 西安电子科技大学学报, 2021, 48(5): 86-91.

KONG Yueping,LIU Chu,ZHU Xudong. Faceanti-spoofing method using the optical flow features of back ground[J]. Journal of Xidian University, 2021, 48(5): 86-91.

图/表 10

图1

图2

图3

图4

图5

图6

表1

图7

图8

表2

参考文献 16

[1]	CHINGOVSKA I, ANJOS A, MARCEL S. On the Effectiveness of Local Binary Patterns in Face Anti-Spoofing[C]// Proceedings of the International Conference of Biometrics Special Interest Group.Piscataway:IEEE, 2012:1-7.
[2]	CAI R, CHEN C. Learning Deep Forest with Multi-scale Local Binary Pattern Features for Face Anti-Spoofing[J/OL]. [2020-03-05]. https://arxiv.org/abs/1910.03850.
[3]	孔月萍, 刘霞, 谢心谦, 等. 基于梯度方向直方图的人脸活体检测方法[J]. 激光与光电子学进展, 2018, 55(03):237-243.
	KONG Yueping, LIU Xia, XIE Xinqian, et al. Face Liveness Detection Method Based on Histogram of Oriented Gradient[J]. Laser & Optoelectronics Progress, 2018, 55(03):237-243.
[4]	李素梅, 秦龙斌, 胡佳洁. 基于特定环境的面部活体检测[J]. 华中科技大学学报(自然科学版), 2018, 46(11):87-91.
	LI Sumei, QIN Longbin, HU Jiajie. Face Liveness Detection System Based on Specific Environment[J]. Journal of Huazhong University of Science and Technology(Natural Science Edition), 2018, 46(11):87-91.
[5]	LINN P P P, HTOON E C. Face Anti-spoofing Using Eyes Movement and CNN-based Liveness Detection[C]// Proceedings of the 2019 International Conference on Advanced Information Technologies.Piscataway:IEEE, 2019:149-154.
[6]	YANG J, LEI Z, LI S Z. Learn Convolutional Neural Network for Face Anti-Spoofing[J/OL]. [2019-09-03]. https://arxiv.org/pdf/1408.5601.pdf.
[7]	LI L, XIA Z Q, HADID A, et al. Replayed Video Attack Detection Based on Motion Blur Analysis[J]. IEEE Transactions on Information Forensics and Security, 2019, 14(9):2246-2261. doi: 10.1109/TIFS.10206
[8]	WU H, ZHAO R, GAN X, et al. Measuring Surface Velocity of Water Flow by Dense Optical Flow Method[J]. Water, 2019, 11(11):2320. doi: 10.3390/w11112320
[9]	CHAUDHRY R, RAVICHANDRAN A, HAGER G, et al. Histograms of Oriented Optical Flow and Binet-Cauchy Kernels on Nonlinear Dynamical Systems for the Recognition of Human Actions[C]// Proceedings of the 2009 IEEE Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE, 2009:1932-1939.
[10]	ANJOS A, CHAKKA M M, MARCEL S. Motion-Based Counter-Measures to Photo Attacks in Face Recognition[J]. IET Biometrics, 2013, 3(3):147-158. doi: 10.1049/bme2.v3.3
[11]	魏冬梅, 周卫东. 近邻样本协作表示的人脸识别算法[J]. 西安电子科技大学学报, 2015, 42(03):115-121.
	WEI Dongmei, ZHOU Weidong. FaceRecognition Using Collaborative Representation with Neighbors[J]. Journal of Xidian University, 2015, 42(03):115-121.
[12]	ZHANG Z W, YAN JJ, LIU S F, et al. A face Antispoofing Database with Diverse Attacks[C]// Proceedings of the 2012 5th IAPR International Conference on Biometrics.Piscataway:IEEE, 2012:26-31.
[13]	PATEL K, HAN H, JAIN A K, et al. Live Face Video vs.Spoof Face Video:Use of Moiré Patterns to Detect Replay Video Attacks[C]// Proceedings of the 2015 International Conference on Biometrics.Piscataway:IEEE, 2015:98-105.
[14]	GAN J Y, LI S L, ZHAI Y K, et al. 3D Convolutional Neural Network Based on Face Anti-Spoofing[C]// Proceedings of the 2017 2nd International Conference on Multimedia and Image Processing.Piscataway:IEEE, 2017:1-5.
[15]	PEREIRA T D F, ANJOS A, DE MARTINO J M, et al. Can Face Anti-spoofing Countermeasures Work in a Real World Scenario?[C]// Proceedings of the 2013 International Conference on Biometrics.Piscataway:IEEE, 2013:1-8.
[16]	ANJOS A, MARCEL S. Counter-measures to Photo Attacks in Face Recognition:a Public Database and a Baseline[C]// Proceedings of the 2011 International Joint Conference on Biometrics.Piscataway:IEEE, 2011:1-7.

实验数据集	E	H	Z
Replay-Attack	2.28	1.25	97.87
CASIA-FASD	6.58	6.04	90.95

方法	ReplayAttack数据集			CASIA-FASD数据集
方法	E	H		E		H
LBP^[1]	14.41		15.45		24.63		23.19
Moire Pattern^[13]			3.30				0.00
CNN^[6]	4.14		2.53		6.98		6.99
3DCNN^[14]	0.16		0.04		5.25		11.34
Motion^[15,16]	11.66		11.79		26.65		30.33
新方法	2.28		1.25		6.58		6.04