Finger Vein Recognition Algorithm Research And System Implementation

Under the background of today's global informatization,the issue of information security has increasingly become a global issue.And the security and confidentiality of information has attracted more and more attention in the information age.Traditional identity verification can easily lead to the loss of personal information and is easy to be stolen.Compared with the biometrics currently used,the finger veins are under the skin of the fingers,and not easy to be stolen.and the finger veins have become an important research direction of biometric technology due to their advantages such as vitality.Deep learning has been gradually applied to the field of biometric recognition due to its powerful feature expression capabilities.Therefore,finger vein recognition based on deep learning has become a technology with the most research significance and practical value.This paper takes self-collected finger vein images as the research object.It aims to construct a deep learning model to extract the veins of finger veins.For the extracted veins,we use a deep convolutional neural network to extract the feature vectors of veins,and construct a target vein comparison algorithm to achieve the identification of target vein.The main research work of this paper are as follows:(1)Propose a finger vein pattern extraction algorithm based on deep learning.On the basis of the U-net network,the deep residual network is used to replace the feature extraction part of U-net.Considering that the pooling layer will lose part of the vein information in the vein extraction of finger veins,a method of replacing traditional convolution with hole convolution is proposed,which can increase the receptive field without pooling,and can better extract the vein informations of finger veins.In order to further improve the expression ability of the model,we use Mish activation function instead of Re LU activation function to make the extracted vein patterns more continuous.Experimental results show that this method can segmente vein patterns well on self-collected data sets.(2)The Face Net-based vein feature vector extraction network is used to extract 128-dimensional feature vectors that can represent the original finger veins.In some image recognition tasks,image classification is used to complete the recognition,but this method is not suitable for data changes.In this paper,the vein feature vector extraction network based on Face Net is used to extract the 128-dimensional feature vector of finger veins,and learns an encoding method from vein image to Euclidean space in an end-to-end manner,and uses Triplet Loss to optimize the model.Furthermore,a target vein comparison algorithm is constructed based on the feature vector of finger veins,where each 128-dimensional vector can be regarded as a point in the 128-dimensional space,and each finger vein image in the data set can be found in the 128-dimensional space.The comparison of the target vein can be completed by comparing the distance of the vector in the space.(3)Based on the above-mentioned vein extraction algorithm and comparison algorithm,a finger vein recognition system is designed to realize the function of system login and user registration.In addition,it also realizes the function modules of user registration vein,vein extraction and vein-based personal identification.