安全指纹识别系统的研究与设计  

Research and Design of Secure Fingerprint Recognition System

作　　者：赵伟舟[1] 杨宝珍[1] 景慧丽[1] 

机构地区：[1]第二炮兵工程大学数学与军事运筹教研室,陕西西安710025

出　　处：《南昌大学学报（工科版）》2012年第3期291-294,共4页Journal of Nanchang University(Engineering & Technology)

基　　金：国家自然科学基金资助项目(60272022)

摘　　要：针对常见的指纹特征提取存在问题以及原始指纹特征的安全问题,进行了安全指纹识别系统的研究与设计。提取了指纹图像感兴趣区域的Frobenius范数特征,借助随机矩阵对范数特征进行映射以增大类间差;并基于映射结果采用阈值化方法生成了作为特征模板的描述子,并将其扩展成为具有安全性和鲁棒性的伪描述子存于系统数据库;对伪描述子采用恢复算法与特征模板进行匹配并给出识别方法。该识别系统中的描述子和伪描述子,都不能通过逆变换获得合法用户的原始指纹特征。FVC2002数据库的实验表明:该识别系统安全、可行、有效,并获得了满意的等错率(0.0219)。Aiming at the problems of common fingerprint feature extraction and security of original fingerprint feature,a secure fingerprint recognition system is designed in this paper.Firstly,Frobenius norm is extracted and mapped by random matrix in ROI,which could improve between-class difference.Descriptor could be obtained by threshold method for the results of random mapping and could be extended to pseudo-descriptor with security and robustness.Recognition method could be got by the recovery algorithm between template and pseudo-descriptor.Neither descriptor nor pseudo-descriptor could recover the original fingerprint feature by the inverse transform.Experiments on FVC2002 showed that the proposed fingerprint-based recognition system is secure,feasible and effective.A satisfactory EER（equal error rate,0.0219） could be got in this paper.

关 键 词：FROBENIUS范数 类间差 描述子 伪描述子 等错率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]