基于深度学习的污损指纹识别研究

发布时间：2018-05-26 07:41

本文选题：指纹识别 + 卷积神经网络　；参考：《电子与信息学报》2017年07期

【摘要】：随着社会信息化水平的提高及不稳定因素的增加,人们迫切需要更加可靠的识别技术对身份进行认证。因此,利用生物特征进行鉴定已成为时下热潮。其中的指纹识别更是因其方便性和可靠性受到普遍认同。传统的指纹识别方法基于特征点比对寻求相似性,此种方法特征点寻找容易出错,且随着指纹的模糊、破坏、污损或是其他问题,均会使识别率明显降低。针对这些问题,该文提出基于深度卷积神经网络(CNN)的CBF-FFPF(Central Block Fingerprint and Fuzzy Feature Points Fingerprint)算法对污损指纹图像进行分类识别。CBF-FFPF算法提取指纹中心点分块图像及特征点模糊化图,合并后输入CNN网络,进行指纹深层特征识别。将该算法与基于主成分分析(KPCA),超限学习机(ELM)和k近邻分类器(KNN)的指纹识别算法进行比较,实验结果表明,所提出的CBF-FFPF算法对污损指纹识别有更高的识别率和更好的鲁棒性。
[Abstract]:With the improvement of the level of social information and the increase of unstable factors, people urgently need more reliable identification technology to authenticate identity. Therefore, the use of biometric identification has become a hot trend. Fingerprint identification is generally accepted because of its convenience and reliability. Traditional fingerprint identification methods seek similarity based on feature point comparison. This method is easy to make mistakes, and with the ambiguity, destruction, defacement or other problems of fingerprint, the recognition rate will be significantly reduced. In order to solve these problems, this paper proposes a CBF-FFPF(Central Block Fingerprint and Fuzzy Feature Points Fingerprint) algorithm based on deep convolution neural network to classify and identify the polluted fingerprint image. CBF-FFPF algorithm extracts the fingerprint center image and the feature point fuzzy image. After merging, the CNN network is input to identify the deep features of fingerprint. The proposed algorithm is compared with the fingerprint identification algorithm based on the principal component analysis (PCA) and the k-nearest neighbor classifier (KNN). The experimental results show that the proposed CBF-FFPF algorithm has higher recognition rate and better robustness.
【作者单位】：杭州电子科技大学网络空间安全学院;杭州电子科技大学通信工程学院;
【基金】：国家重点研发计划(2016YFB0800201) 浙江省自然科学基金(LY16F020016) 浙江省重点科技创新团队项目(2013TD03)~~
【分类号】：TP18;TP391.41

【相似文献】