Finger Vein Recognition Research Based On Deep Learning

With the rapid development of Internet,people's life has become more convenient.Internet technology brings us convenience as well as challenges to information security of citizens.Therefore,it is of great significance to study efficient and safe biometric authentication technology.Compared with human face recognition and fingerprint identification,the finger vein recognition is more safe and effective.A finger vein recognition algorithm based on depth learning is proposed in this paper.The main work of this paper is as follows:1.The preprocessing of the original image collected by the finger vein was analyzed.In the acquisition process of the finger vein image,the finger will rotate at random angle.In this paper,an angle correction algorithm is proposed to correct the angle rotation of the finger.The venous image contains a large number of useless background areas,and the difference in the intensity of light makes the outline of the finger blurred.In order to obtain a good area of the finger vein,it is necessary to detect the finger boundary accurately.In this paper,the vertical edge detection and morphological algorithm are used to detect the finger boundary,and the burr removal algorithm is proposed to solve the problem of boundary line burr.2.Paper analyzes optimization algorithm of venous image.Matching filter and morphological method are affected by the degree of vein thickness and brightness.In this paper,we use curvature concept to optimization vein features.The use of curvature to analyze the curve of the finger section can reduce the influence of the degree of vein thickness and brightness.3.The algorithm of finger vein recognition research based on deep learning is proposed,AlexNet convolutional neural network performs well in image classification.In this paper,we deeply analyze the structure of the network model.In order to reduce the number of network model parameters,we improve the network and train the network model in the framework of Caffe.The study shows that the improved depth neural network can be used to classify the extracted venous feature data better.