摘要
基于图模型的指静脉全局特征表达方法不仅可以降低成像质量对采集设备的依赖性,还能提高匹配效率。针对于目前指静脉图模型的研究中存在的图结构不稳定,匹配效率随图模型的变大而降低的问题,本文提出了一种基于SLIC(Simple Linear Iterative Clustering)超像素分割算法构建加权图的方法,并改进ChebyNet图卷积神经网络(Graph Convolutional Neural Networks,GCNs)提取加权图的图级(graph-level)特征。针对指静脉样本数普遍较少,而ChebyNet中卷积网络参数量较大容易造成过拟合以及其快速池化层不能自适应地选择节点的问题,本文提出了全局池化结构的改进GCNs模型SCheby-MgPool(Simplified Cheby-Multi gPool)。实验结果表明,本文提出的方法提取的指静脉特征在识别精度,匹配效率上都具有较好的性能。
The global feature representation method of finger-vein based on graph model can not only reduce the dependence of imaging quality on acquisition equipment,but also improve the matching efficiency.Aiming at the problems of unstable graph structure in the current study of finger-vein graph models,that with the matching efficiency decreases as the graph model becomes larger,this paper proposed a method of constructing weighted graph based on SLIC(Simple Linear Iterative Clustering)superpixels segmentation algorithm,and improved ChebyNet graph convolutional neural networks(GCNs)to extract graph-level features of weighted graph.Because the number of finger-vein samples is generally small,while the number of parameters in ChebyNet are large,it is easy to cause over-fitting and the problem that its fast pooling layer unable to adaptively select the nodes.In this paper,an improved GCNs model SCheby-MgPool(Simplified Cheby-Multi gPool)with global pooling structure was proposed.The experimental results show that the finger-vein features extracted according to the method proposed in this paper have better performance in recognition accuracy and matching efficiency.
作者
李冉
苏志刚
张海刚
杨金锋
Li Ran;Su Zhigang;Zhang Haigang;Yang Jinfeng(Tianjin Key Laboratory for Advanced Signal Processing,Civil Aviation University of China, Tianjin 300300, China;Sino-European Institute of Aviation Engineering,Civil Aviation University of China, Tianjin 300300, China;Institute of Applied Artificial Intelligence of the Guangdong-Hong Kong-Macao Greater Bay Area, Shenzhen Polytechnic, Shenzhen, Guangdong 518055, China)
出处
《信号处理》
CSCD
北大核心
2020年第4期550-561,共12页
Journal of Signal Processing
基金
国家自然科学基金(61806208)。
关键词
指静脉识别
加权图
特征表达
图卷积神经网络
finger-vein recognition
weighted graph
feature representation
graph convolution neural networks
作者简介
李冉,女,1996年生,河南商丘人,中国民航大学电子信息与自动化学院硕士研究生,主要研究方向为图像处理、生物特征识别,E-mail:2018022104@cauc.edu.cn;通信作者:苏志刚,男,1972年生,黑龙江尚志人,中国民航大学中欧航空工程师学院,教授,博士,主要研究方向为信号与信息处理、及其在监视与导航领域的应用研究,E-mail:ssrsu@vip.sina.com;张海刚,男,1989年生,河北沧州人,深圳职业技术学院,讲师,博士,主要研究方向为机器学习、图像处理、生物特征识别,E-mail:zhg2018@sina.com;杨金锋,男,1971年生,河南郑州人,深圳职业技术学院,教授,博士,主要研究方向为模式识别、计算机视觉,E-mail:jfyang@szpt.edu.cn。
