基于深度学习的指纹自动识别技术  

Introduction to Automatic Fingerprint Identification Based on Deep Learning

作　　者：吴春生 李孝君[2] 吴浩[2] WU Chunsheng;LI Xiaojun;WU Hao(Beijing Police College,Beijing 102202,China;Institute of Forensic Science,Ministry of Public Security,Beijing 100038,China)

机构地区：[1]北京警察学院,北京102202 [2]公安部物证鉴定中心,北京100038

出　　处：《刑事技术》2022年第1期88-95,共8页Forensic Science and Technology

基　　金：中央级公益性科研院所基本科研业务费专项资金项目(2019JB003);痕迹科学与技术公安部重点实验室开放课题(2019FMKFKT01)。

摘　　要：本文从学科领域入手,对指纹自动识别技术在发展过程中受人工智能技术影响所产生的新变化进行简述。指纹识别技术作为一种计算机应用技术,其发展与计算机科学的新技术密切相关。人工智能技术,特别是基于深度学习的图像技术的发展使指纹识别算法开启了全新的模式。本文将人工智能在指纹领域的发展分成三个阶段,并对当前所处的第二阶段的发展趋势进行了分析。基于深度学习的指纹识别技术使用图像特征而不是传统细节点特征的方式,改变了法庭科学领域对指纹识别的认知。本文重点对深度学习技术在指纹识别方面的应用模式和典型的技术方法进行了论述,给出了基于深度学习的指纹识别技术方案图,对技术方案中的网络模型设计等重要步骤逐一进行了说明,提出了图像处理、降维等几个需要重点攻坚的技术环节。对现有的可为指纹识别借鉴使用的深度网络模型进行了介绍,如:卷积神经网络、自编码器网络。最后对人工智能指纹识别算法与传统算法的性能进行了对比。An introduction was here put forward about automatic fingerprint identification relating to deep learning.Starting from the brief review on automatic fingerprint identification technology(i.e.,the automated fingerprint identification system,AFIS)and developing process,the computer-based innovations were further elucidated to focus on artificial intelligence(AI)technology and its application in recognition of images.Fingerprint identification is definitely a kind of image discrimination,hence having it brought into a new way of AI technology tackling.With the recognition algorithm about fingerprint pattern,AI is playing its role in fingerprint identification through improving image technology based on deep learning.AI’s fingerprint identification can be divided into three stages,leaving the second stage being here specifically described about the relevant developing trend.Forensic fingerprint recognition has since been changed with deep learning adopting image features other than traditional minutiae distinctions.The emphasis of this essay was put on the deep learning technology about its application mode and typical technology method in fingerprint identification,with the demonstrative technical scheme being illustrated.For the involving network model design in the technical scheme mentioned above,those important steps were explained one by one.Several key technical problems were also proposed about image processing,dimension reduction and the others related.The present available deep learning network models,eligible for fingerprint identification,were introduced with the examples of convolutional neural network and the auto-encoder one,too.Finally,the comparison was carried out on performance of AI’s fingerprint identification algorithm against the traditional one.

关 键 词：深度学习 人工智能 指纹识别 无特征比对 指纹自动识别系统 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]