The biometric recognition on contactless multi-spectrum finger images
The biometric recognition on contactless multi-spectrum finger images
复制标题
非接触式多光谱手指图像生物特征识别
DOI：
10.1016/j.infrared.2014.10.007
复制
发表时间：
2015-01-01
影响因子：
3.3
通讯作者：
Wu, Qiuxia
中科院分区：
文献类型：
--
作者：
Kang, Wenxiong;Chen, Xiaopeng;Wu, Qiuxia
针对单模态生物特征识别的局限性，提出了一种基于非接触式多光谱指纹图像的多模态生物特征识别系统。该系统的主要优点是允许纹理的丰富性和数据访问的方便性。我们构建了一个多光谱仪器，可以同时从手指上获取三种不同类型的生物特征：非接触式指纹，手指静脉和指关节纹。在这些特征样本的基础上，建立了一个适度的数据库，以评估我们的系统。考虑到实时性要求和三种生物特征的各自特点，指纹和指静脉采用分块局部二值模式算法进行特征提取和匹配，指关节纹采用定向FAST和旋转BRIEF算法。最后，对上述三种生物特征的匹配结果进行评分级融合。实验结果表明，我们提出的多模态生物特征识别系统达到了0.109%的等错误率，这是88.9%，94.6%，89.7%，低于个人指纹，指关节纹，手指静脉识别，分别。尽管如此，我们提出的系统也满足应用程序的实时性要求。(C)2014爱思唯尔有限公司版权所有。
This paper presents a novel multimodal biometric system based on contactless multi-spectrum finger images, which aims to deal with the limitations of unimodal biometrics. The chief merits of the system are the richness of the permissible texture and the ease of data access. We constructed a multi-spectrum instrument to simultaneously acquire three different types of biometrics from a finger: contactless fingerprint, finger vein, and knuckleprint. On the basis of the samples with these characteristics, a moderate database was built for the evaluation of our system. Considering the real-time requirements and the respective characteristics of the three biometrics, the block local binary patterns algorithm was used to extract features and match for the fingerprints and finger veins, while the Oriented FAST and Rotated BRIEF algorithm was applied for knuckleprints. Finally, score-level fusion was performed on the matching results from the aforementioned three types of biometrics. The experiments showed that our proposed multimodal biometric recognition system achieves an equal error rate of 0.109%, which is 88.9%, 94.6%, and 89.7% lower than the individual fingerprint, knuckleprint, and finger vein recognitions, respectively. Nevertheless, our proposed system also satisfies the real-time requirements of the applications. (C) 2014 Elsevier B.V. All rights reserved.