基于图卷积网络的手指静脉识别方法研究

针对传统手指静脉识别方法往往存在识别率低或者计算量大等问题，本文提出一种基于轻量型图卷积网络的手指静脉识别方法。首先用一个加权图描述一张手指静脉图像，图的顶点特征和加权边集分别由指静脉图像的局部方向能量特征和特征间相关性确定。图数据作为输入，经过基于切比雪夫多项式的图卷积层和由图粗化协助的快速池化层，然后全连接层进行特征整合，再进行分类识别。实验结果显示，该方法识别效率远高于传统算法，并在实验室自制手指静脉数据库达到96.80%的识别率，在不同数据库有较好的普适性。 

Finger-vein Recognition Based on Graph Convolutional Networks

A new finger-vein recognition method based on the lightweight Graph Convolutional Network was proposed to solve the problems that the traditional finger-vein recognition methods often had a low accuracy or a huge computational complexity.Before the graph convolution,a finger-vein image was expressed as a weighted graph.The graph’s nodes were decided by the feature of the finger-vein image’s local orientated energy and the correlation between nodes’feature determined the edges’weight.As input data,the graph was learned by graph convolution layer defined by Chebyshev polynomial,and fast pooling layer assisted by graph coarsening.Then full connected layers were added to achieve the recognition of finger-vein graph.The experiment results show that the recognition efficiency is much higher than the traditional algorithm and the recognition accuracy is 96.80%in our homemade finger-vein database.Meanwhile,the university of the method is proved in different database.