基于CNN掌纹识别技术在智慧电网的应用研究  

Research on the application of CNN palmprint recognition technology in smart power grid

作　　者：陈虹旭 孙怡然 李晓坤 刘清源 徐龙 董潍赫 CHEN Hongxu;SUN Yiran;LI Xiaokun;LIU Qingyuan;XU Long;DONG Weihe(Postdoctoral Program of Heilongjiang Hengxun Technology Co.,Ltd.,Harbin 150090,China;College of International Culture and Education Heilongjiang University,Harbin 150090,China)

机构地区：[1]黑龙江恒讯科技有限公司国家博士后科研工作站,哈尔滨150090 [2]黑龙江大学,哈尔滨150090

出　　处：《智能计算机与应用》2020年第6期92-97,共6页Intelligent Computer and Applications

基　　金：中小企业创新基金(2017FF1GJ023);专利优势示范企业基金(2017YBQCZ029);国家自然科学基金(81273649);国家自然科学基金(61501132)中央高校基本科研业务费专项资金(3072019CFT0603);国家自然科学基金(61672181);黑龙江省自然科学联合引导基金(LH2019F049,LH2019A029);中国博士后科学基金(2019M650069);黑龙江省基础科研科技创新基金(KJCX201805);黑龙江省基础科研青年创新团队基金(RCYJTD201805)。

摘　　要：目前智慧电网的信息安全隐患较大,本文提出一种基于CNN的掌纹识别模型,如AlexNet,ResNet等,将其与智慧电网相融合,可以改善或解决智慧电网信息系统的诸多安全问题。生物特征识别被认为是一种强大而有效的监视和安全应用技术。具有代表性的生物特征包括掌纹,指纹,虹膜等,不仅包含有效且稳定的特征,还包含丰富的纹理特征,引起了很多关注。由于其低成本、用户友好和强健的属性,基于掌纹的识别已逐渐应用于许多民用应用。深度学习方法被认为是计算机视觉领域的一大突破,在包括生物特征识别在内的许多领域都得到了成功的应用,掌纹识别以较高的可接受性被接受。本研究将深度学习方法,卷积神经网络引入掌纹识别中,利用Hausdorff距离来匹配特征向量,以获得更好的识别效果。实验结果表明,与传统的识别方法(如PCA、LBP)相比,基于卷积神经网络的掌纹识别率更高。Since the information security risks of smart grid are large at present,we propose a palm print recognition model based on CNN,such as AlexNet,ResNet,etc.By integrating it with smart grid,we can improve and solve many security problems of information system of smart grid.Biometrics is considered to be a powerful and effective monitoring and security application technology.Representative biometrics,including palmprint,fingerprint,iris,etc.,not only contain effective and stable features,but also contain rich texture features,which have attracted a lot of attention.Due to its low cost,user-friendly and robust properties,palmprint recognition has been gradually applied to many civilian applications.Deep learning is considered as a breakthrough in the field of computer vision and has been successfully applied in many fields including biometric recognition.Palmprint recognition is accepted with high acceptability.In this study,deep learning method and convolutional neural network are introduced into palmprint recognition and Hausdorff distance is used to match feature vectors,so as to obtain better recognition effect.Experimental results show that,compared with the traditional recognition methods(such as PCA and LBP),the recognition rate of palmprint based on convolutional neural network is higher.

关 键 词：掌纹识别 CNN Alexnet HAUSDORFF距离 智慧电网 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]