基于Capsule Net架构的模糊指纹图像识别  

Fuzzy Fingerprint Image Recognition Based on Capsule Net

作　　者：袁颖 李论[2] 杨英仓 YUAN Ying;LI Lun;YANG Ying-cang(Guizhou Police College,Guiyang 550005,China;Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州警察学院,贵州贵阳550005 [2]贵州大学,贵州贵阳550025

出　　处：《软件》2020年第8期48-51,共4页Software

基　　金：贵州省科技厅基础研究计划项目(黔科合基础[2017]1063)。

摘　　要：指纹识别是目前应用最广泛的生物特征识别技术之一,传统指纹识别技术中为了使原始指纹图像特征更易于被当前的计算方法处理,需要对指纹图像特征点标画,为输入数据设计好表示层,再仿射变换到另一个表示空间,难以得到模糊指纹图像所需要的精确表示。本文采用胶囊神经网络,使用端到端的深度学习模型来完成模糊指纹图像的识别,克服了传统机器学习算法对于感知问题不敏感,以及卷积神经网络缺乏整体特征的明确信息和最大池化中丢失了有价值的特征信息等劣势。实验结果表明,胶囊神经网络在模糊指纹图像处理中具有较高的识别精度。Fingerprint identification is currently the most widely used biological characteristic.In traditional fingerprint identification technology,in order to make the fingerprint image feature calculate more easier,method have to sign the fingerprint image feature points,input data design good presentation layer,and affine transformation to another representation space to realize recognition.However this method is hard to get the precise fuzzy fingerprint image features.Capsule neural network in this paper use end-to-end deep learning model to match the fuzzy identification of fingerprint image.This method overcomes the unsensitive problems to perceptual features which is a common problems in the traditional machine learning algorithm.This method also overcomes the lack of clear information and the lost of characteristics of the information in integral pooling in the convolutional neural network.The experimental results show that the capsule neural network has high recognition accuracy in fuzzy fingerprint image processing.

关 键 词：指纹识别 胶囊神经网络 深度学习 

分 类 号：TP391.[自动化与计算机技术—计算机应用技术]