Study On Finger Vein Feature Extraction Method

With the rapid development of economy and technology, personal identity occupies a very important position in all walks of life and was applied more and more widely. In recent years, many kinds of biometric identification technology walk into all aspects of people’s life gradually, including face, fingerprint, vein and iris applied to access control, security, etc. Compared with the traditional password, ID card and key, biometric identification technology is difficult to change, lose and forge.The finger vein recognition technology is an emerging biometric technology and superior to other biometric identification technology, because it is with high technology content and difficult to forge. Since both fingerprints and knuckles lines are on the surface of the skin, easily damaged, this will affect the recognition effect. By contrast, finger veins are in the inside of the skin, protected by skin and not easily damaged. At present, although the market has emerged a lot of more mature application products, there are some problems in recognition rate and speed.This paper proposes a differential geometry algorithm to extract the characteristics of finger vein, combining with the Hausdorff distance matching, achieve the matching function of heterologous finger and homologous finger. Firstly, building hand images with thumb outward open and four fingers closed, segmenting hand from the background with the gray histogram threshold segmentation method, detecting the suture with the local gray minimum method, determining the coordinates of three finger roots. Due to the direction of the hand is not standard, it can affect the feature matching and the identification rate of finger vein recognition system. So the direction normalization process is necessary. The method is connecting two endpoints of lines between the middle finger and the ring finger, rotating the image under the level of the straight line, making the direction of normalization, and determining the coordinates of finger roots. Capture the interest of image, gray equalization processing on ROI, making the finger vein lines more clearly. In this paper, using differential geometry algorithm to extract the finger vein characteristic after the image preprocessing, a window scanning in the vein image point by point, approximating the image by its second or third order Taylor polynomial, the coefficients of this polynomial are usually determined by using the facet model, the direction of the line is determined from the Hessian matrix of the Taylor polynomia, judging center point on the direction of a first order directional derivative is zero. If it is zero, the center point is the vein lines line feature. Acquiring left hand images of 18 people, getting 5 images from the same person, this is the self-built image library. Based on the self-built image library test in this paper, respectively measured reject rate was 18.89% and the course at a rate of 11.71% on homologous finger 1:1 authentication mode and heterologous finger 1: n authentication mode. Meanwhile, recognition rate of the algorithm is around 73% in the embedded environment.