Research On The Algorithm Of Finger Vein Recognition

With the rapid development of the information age,People's demand for information safety is becoming higher,and the traditional identity authentication method can no longer meet people's needs.Biometric recognition technology has been widely concerned by some researchers due to its high security level and convenience.Compared with other biometrics,finger vein recognition technology has become a research hotspot in biometrics owing to its advantages like non-contact,in vivo recognition and internal features.However,there still exist some problems in the finger vein recognition algorithms,such as inaccurate extraction of vein lines,low system recognition performance,and poor real-time performance of the system.In order to improve the performance of the system,this dissertation launches a deep research on the following aspects:To deal with the problems of poor segmentation effect existed in the current segmentation algorithms,a segmentation algorithm based on two-dimensional Gauss curvature is proposed in this dissertation.There are two main innovations in this algorithm: firsrly proposes the combination of curvature and local area dynamic threshold algorithm,and secondly proposes extract the curvature of finger vein image using the two-dimensional Gauss kernel function,on the one hand,to filter,and on the other hand,to weaken the effect of noise by fully using all the pixels in the template region.Theoretical analysis and experimental simulation results show that the proposed segmentation algorithm is more accurate and stable than that of classical maximum curvature algorithm and vein filtering based on directional filtering algorithm.To handle the the low performance of the current system recognition algorithm,this dissertation proposes a recognition algorithm based on the thinning image correction translation and the direction field map.The traditional MHD recognition algorithm based on feature point operates the matching recognition using the feature points,and its performance is low due to thinning image,preudo feature points,and lack of real feature points,but it verifies the stability and uniqueness of the refined images.Thus,this dissertation proposes to use thinning images to correct translation between images.The direction field map is the quantized value determined by the relative measurement of different direction of gray value difference,and each pixel in the direction of the field map contains the image information,which has stability and integrity.Thus,this dissertation proposes a recognition algorithm based on the direction field map.Simulation results show that the thinning image can correct the translation between images,to some extent,and the performance of the recognition algorithm based on the direction field map is much better than the MHD recognition algorithm based on the feature points and the gradient correlation coefficient recognition algorithm.The finger vein recognition system has two models of discrimination and verification,that is,1:1 mode and 1:N mode.In 1: N mode,with the increase of N,the time-consuming of the matching recognition increases linearly.In this dissertation,a fast recognition algorithm based on step thinking is proposed to reduce the time of matching recognition by using the thinned image and direction field map.Firstly,we use the thinning image of the finger vein image to distinguish the same kind of images,and then use the direction field map recognition algorithm to further operate the matching recognition.Theoretical analysis and experimental simulation results show that the fast recognition algorithm based on the thinned image and direction field map can improve the real-time performance of the system.This dissertation has studied the algorithm of finger vein recognition in a comprehensive and systematic way.The segmentation algorithm,recognition algorithm and 1:N fast recognition algorithm proposed in this dissertation improve the reliability and effectiveness of the system,and promote the application and development of finger vein recognition technology to a certain extent.