基于深度学习的指纹方向场提取算法  

Fingerprint orientation field estimation algorithm based on deep learning

作　　者：刘永鸿 韩丛英 郭田德[2,3] LIU Yonghong;HAN Congying;GUOTiande(College of Big Data Statistics,Guizhou University of Finance and Economics,Guiyang 550025,Guizhou,China;School of Mathematical Sciences,University of Chinese Academy of Sciences,Beijing 100049,China;Key Laboratory of Big Data Mining and Knowledge Management,Chinese Academy of Sciences,Beijing 100190,China)

机构地区：[1]贵州财经大学大数据统计学院,贵州贵阳550025 [2]中国科学院大学数学科学学院,北京100049 [3]中国科学院大数据挖掘与知识管理重点实验室,北京100190

出　　处：《运筹学学报》2023年第4期1-19,共19页Operations Research Transactions

基　　金：国家自然科学基金(Nos.U19B2040,11991022);贵州财经大学引进人才启动项目(No.2021YJ005)。

摘　　要：作为指纹图像中的一个非常重要的特征,指纹方向场在自动指纹识别系统的很多环节中扮演着重要的角色,例如指纹图像增强、奇异点提取、指纹分类等。尽管现有的方向场提取算法可以取得不错的提取效果,但是这些算法对于图像噪声比较敏感,同时经常需要先验知识进行方向计算,算法运行也消耗了很多时间。针对指纹方向场提取问题,本文提出了一种基于全卷积网络的方向场提取算法,利用像素级别的分类任务估计方向场。根据指纹图像与注意力机制的特点,设计了一个用于提取方向场的注意力机制的全卷积网络,并在网络中添加了空洞卷积层,有效提取了不同指纹图像中重要的判别特征,同时设计了一个新的损失函数来训练网络,最终根据像素点的分类结果实现了方向场的提取。实验结果表明,本文的算法实现了较好的提取效果以及较快的提取速度,对于图像噪声等具有很好的鲁棒性。As a very important feature in fingerprint images,fingerprint orientation field plays an important role in many aspects of Automatic Fingerprint Identification System(AFIS),such as fingerprint image enhancement,singular point detection,fingerprint classification,etc.Although existing orientation field estimation algorithms can achieve competitive extraction results,these algorithms are sensitive to image noise and often require prior knowledge for direction calculation,which consumes a lot of time.To address this issue,a fingerprint orientation field estimation algorithm based on fully convolutional network is proposed in this paper,which uses pixel-level classification tasks to estimate fingerprint orientation field.According to the characteristics of fingerprint image and attention mechanism,we design a fully convolutional network with attention mechanism(Attention FCN)for orientation field estimation.In addition,dilated convolution layers are added to the network to extract important discriminative features in different fingerprint images.At the same time,a new loss function is designed to train the network.According to the classification results of each pixel point,the orientation field is estimated.Experimental results show that the proposed algorithm achieves good performance and very fast estimation speed,and it is very robust to image noise.

关 键 词：指纹方向场提取 深度学习 注意力机制 全卷积网络 

分 类 号：O221.2[理学—运筹学与控制论]