基于注意力残差网络的OCT指纹防伪方法  

Anti-Spoofing Research of OCT Fingerprint Based on Attention Residual Network

作　　者：王海霞 潘栋 梁荣华[1] 张怡龙 刘义鹏 陈朋 Wang Haixia;Pan Dong;Liang Ronghua;Zhang Yilong;Liu Yipeng;Chen Peng(College of Computer Science and Technology,Zhejiang University of Technology,Hangzhou 310023;College of Information Engineering,Zhejiang University of Technology,Hangzhou 310023)

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310023 [2]浙江工业大学信息工程学院,杭州310023

出　　处：《计算机辅助设计与图形学学报》2022年第3期392-402,共11页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(61976189,61905218,62076220,61527808);浙江省自然科学基金(LY19F050011);浙江省属高校基本科研业务费专项(RF-C2019001)。

摘　　要：基于光学相干断层扫描(optical coherence tomography, OCT)的指纹防伪主要依赖汗孔、皮下汗腺和内部指纹等信息的精确提取,且具有一定的防伪信息冗余.针对这一现状,提出了一种基于注意力残差网络的OCT指纹防伪方法,仅使用少量信息便可做到对指纹的真伪识别.首先在残差网络的基础上,通过注意力模块与长短跳跃连接,构建注意力残差网络,使网络注意力更好地集中在皮下组织信息上;其次对待判别手指体数据进行随机采样,提取局部特征小块;最后通过注意力残差网络对局部特征小块进行分类,实现手指的真伪识别.基于Keras框架的注意力残差网络方法,在自制真假指纹数据库中,较同类网络获得了更高的识别准确率,表明了所提防伪方法的优越性.对自制数据库和深圳大学OCT指纹数据库的跨设备、跨样本实验进一步验证了方法的准确性和鲁棒性.Current optical coherence tomography(OCT) based fingerprint anti-counterfeiting methods are mainly through the extraction of sweat pores,subcutaneous sweat glands and internal fingerprints,which depend on the extraction accuracy and contain redundant anti-counterfeiting information.An OCT fingerprint anti-counterfeiting method based on the attention residual network is proposed,where only a small amount of data is required to identify the authenticity of the fingerprint.Firstly,the residual network is incorporated with the convolutional block attention module as well as the long and short skip connections to better focus the network attention on the subcutaneous tissue information.Secondly,random sampling is performed on the volume data of the discriminating finger,and the local feature patches are extracted.In the last,these local feature patches are classified by the attention residual network to realize the identification of the true and fake fingerprint.Experiments are carried out using the Keras framework and on the self-established true and fake fingerprint database.The proposed method obtains a higher anti-counterfeiting accuracy rate than other networks,which proves the superiority of the proposed method.The cross-device and cross-sample experiment on self-established database and Shenzhen University OCT fingerprint database has further verified the accuracy and robustness of the proposed method.

关 键 词：光学相干断层扫描 指纹防伪 残差网络 卷积注意力模块 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]