The Research Of Finger Vein Recognition

Finger vein recognition technology refers to the use of finger vein angiography image authentication technology,which has became one of the hot spots of biometric identification technique.This thesis has studied the feature extraction and recognition process in finger vein recognition.The main contents include:(1)The Modified Entropy Component algorithm is proposed to modify the ECA,the separation matrix W is introduced into the feature space of the Entropy Component Analysis method based on the principle of negative entropy maximization in the Infomax algorithm.It is used to extract the optimized entropy components,and further extract the more favorable classification characteristics.(2)2D Modified Entropy Component Analysis algorithm is proposed to further ultimate the contruct information,which performance better than Modified Entropy Component algorithm.2D Gabor filter's real part can effectively analyze the texture of multi-scale and multi-directional of the objection.Based on a fusion rule of the 2D Gabor feature and 2D Modified Entropy Component algorithm,the finger vein recognition rate can be further evaluated.Also,we propose a strategy of the fusion 2D Gabor feature combining with Blocked Local Binary Pattern algorithm to strength the rotation invariance and grayscale invariance.(3)The effect of the depth learning model on image recognition is more effective than the traditional machine learning method,but at the same time,the deep learning network requires a large number of training samples,and the equipment complexity is high.The deep learning network has lots of hyperparameters,the tuning of the hyperparameters is a very complicated project.Deep forest is the application of deep learning model to improve random forest algorithm,which open a new field of deep learning.The deep forest is applied to the finger vein recognition and combined with LBP response images,block LBP histogram features.The deep forest proposed a new method for finger vein recognition,experiments show that the deep forest based finger vein recognition method compared with the traditional identification method has a higher recognition rate.