基于卷积神经网络的多光谱掌纹识别技术  

Multi-spectral Palmprint Recognition Technology Based on Convolutional Neural Network

作　　者：刘雪微 王磊[2] 章强[2] 王继帅[2] 李选普 LIU Xuewei;WANG Lei;ZHANG Qiang;WANG Jishuai;LI Xuanpu(School of Information Science and Engineering,Harbin Institute of Technology(Weihai),Weihai 264209,China;Suzhou Institute of Biomedical Engineering and Technology,Chinese Academy of Sciences,Suzhou 215163,China;School of Automation,Central South University,Changsha 410083,China)

机构地区：[1]哈尔滨工业大学(威海)信息科学与工程学院,山东威海264209 [2]中国科学院苏州生物医学工程技术研究所,江苏苏州215163 [3]中南大学自动化学院,湖南长沙410083

出　　处：《郑州大学学报（理学版）》2021年第3期50-55,共6页Journal of Zhengzhou University:Natural Science Edition

基　　金：国家重点研发计划项目(2019YFC0121800);中国科学院科研仪器设备研制项目(YJKYYQ20170068)。

摘　　要：在多光谱掌纹图像采集设备研发的基础上,提出一种基于卷积神经网络的多光谱掌纹识别方法。该方法使用ResNet-18网络对掌纹图像进行特征提取,应用特征级融合方法并将融合后的特征送入SVM分类器中进行决策。实验结果表明,所提出的多光谱掌纹识别方法可以有效提升掌纹图像特征判别效率,相比于单光谱方法和传统的神经网络算法具有更高的识别准确率。Based on the development of multi-spectral palmprint image acquisition equipment,a multispectral palmprint recognition method based on convolution neural network was proposed.The ResNet-18 network was used to extract the features of palmprint image,then the feature-level fusion algorithm was applied and the fused features were sent to SVM classifier for decision-making.Experimental results showed that the proposed multi-spectral palmprint recognition method could effectively improve the recognition efficiency of palmprint image features,with higher recognition accuracy than single spectrum method and traditional neural network algorithm.

关 键 词：多光谱融合 卷积神经网络 特征提取 掌纹识别 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]