融合注意力机制和残差网络的掌纹识别  

Palmprint Recognition Integrating Attention Mechanism and Residual Network

作　　者：于霞[1] 付琪 薛丹[1] 王健行 武家逸 赵鑫峰 YU Xia;FU Qi;XUE Dan;WANG Jianxing;WU Jiayi;ZHAO Xinfeng(School of Information Science&Engineering,Shenyang University of Technology,Shenyang 110870,China)

机构地区：[1]沈阳工业大学信息科学与工程学院,辽宁沈阳110870

出　　处：《沈阳大学学报（自然科学版）》2023年第6期511-520,共10页Journal of Shenyang University：Natural Science

基　　金：国家自然科学基金资助项目(62301339);辽宁省教育厅高等学校基本科研资助项目(LJKMZ20220478)。

摘　　要：为了提高掌纹识别算法的准确率,解决掌纹纹理信息利用率不高的问题,给出了一种融合高效通道注意力机制的改进残差网络(ECA-MNet)对掌纹图像进行分类。在原始残差网络的基础上对残差模块进行改进,将多个改进的残差模块进行拼接,同时在残差模块内部添加高效通道注意力机制模块,通过权重分配来突出重点特征。实验结果表明,ECA-MNet在4种公开掌纹数据集上的识别准确率较其他经典网络模型均有提升,在自建掌纹数据集上的识别准确率达到98.21%。In order to improve the accuracy of the palmprint recognition algorithm and solve the problem of low utilization of palmprint texture information,an improved residual network(ECA-MNet)based on efficient channel attention mechanism was proposed to classify palmprint images.The residual module was improved on the basis of the original residual network.Multiple improved residual modules were spliced,and efficient channel attention mechanism modules were added inside the residual modules to highlight key features through weight allocation.The experimental results showed that the recognition accuracy of ECA-MNet on the four public palmprint datasets was improved compared with other classical network models,and the recognition accuracy on the self-built palmprint datasets reached 98.21%.

关 键 词：掌纹识别 深度学习 图像处理 注意力机制 残差网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]