Research On Feature Extraction Algorithm In Finger Vein Recognition

In modem society,with the rapid development of information technology and the continuous extend of human physical and virtual space,human identity recognition is be put forward higher requirements for safety and practicality.Among various kinds of biometric identifiers,finger vein recognition is a new biometric technology which have some better prospects with non-contact?high security and live body recognition.So,the identity recognition based on biometrics has attracted more and more researchers' attention.In this paper,from the reality of finger vein recognition technology,the texture of finger vein image as the center,multi-neighborhood pixels of correlate and the neighborhood feature bit of robust as basic point to depth research.And main works are shown in following:In the finger image positioning part,this paper uses a sliding window to locate the finger image.The method can be divided into three steps:detect and correct the rotated image,the height of located by a sliding window and the width of located by tangent line.The related LBP feature extraction algorithm based on the image texture feature mostly selects the single-ring neighborhood or the square neighborhood as the feature extraction block of the image.Therefore,there is only a single level correlation between the neighborhood sampling point and the central pixel,but considering the growth direction of the finger vein texture,finger vein images in the multi-level neighborhood sampling points should also be corrected,Based on this,this paper proposes a dual binary pattern(DBP)based on finger vein recognition algorithm.Firstly,introduced a two-level extraction block to the ROI image feature encoding,and then feature vector of image is calculated by statistics block histogram,finally use the histogram intersection kernel to match the image.The traditional feature extraction algorithm assigns the same meaning to the"bit" of the binary code,and does not take into account the correlation between the characteristic "bit" and the growth direction of the vein texture.Based on this idea,this thesis proposes a global direction-local binary pattern(GD-LBP)of finger vein recognition technology.Firstly,the circular neighborhood sampling points in the horizontal and vertical direction,oblique direction respectively binary encoding,and then integrating the two sub-direction coding results to improve the local binary pattern of the finger vein feature,lastly,the feature vector of the image is obtained by statistical block histogram method.This paper conducts contrast experiments on SDUMLA-FV databases and ChuangHe databases.The experiment results shows that DBP algorithm makes full use of the correlation between multi-neighborhood pixel sampling points and the growth direction of vein texture,enhances the robustness of image pixel levels,improves the recognition accuracy of the system and reduces the equal error rate of system.The GD-LBP algorithm complements the characteristic advantage of the vein image in the sub-direction and enhances the stability of the image feature "bit",which can be applied to the system with high real-time requirement.