Study On Finger Vein Recognition Methods

In modern society, with the rapid development of computer and network technology, the information security becomes more and more important. People expect a more convenient and reliable method for identify authentication. The biometrics makes this possible. Fingerprint recognition is considered as the comparative maturity technology in biological feature recognition. However, its fatally defect is easy to forge, destroy or contaminate, which would restrict its application. Finger vein conceals body interior, which has many merits such as living body recognition, interior characteristic, non-contacting and high safety class, and so on, that can effectively resolve the many problems in the traditional biometrics.Main jobs are as follows.(1) One kind of finger vein image segmentation algorithm based on fuzzy theory is proposed. It combined traditional fuzzy theory with the classical image segmentation algorithm. Different vein images are applyied to test, experiments proved that it is effective in extracting vein feature by NiBlack threshold method, and enhances precision of feature extraction.(2) A vein segmentation algorithm based on direction and detecting valley region is proposed. The method is achieved by detecting valley region and image segmentation. The experiment results show that it is effective not only in extracting vein pattern but also in removing noise and false vein feature. At the same time, it can improve the recognition precision of the whole system and the speed of image process.(3) The paper proposed a binary image thinning method. The template is obtained by analysing feature of redundant pixels in the vein image thinned with traditional thinning method, and then utilizing template to thin the image further. Fnally, vein binary image with single pixel can be obtained by pruning burr and repair, which is helpful to the following process.(4) An improving and achieving match method based on vein geometry is discussed. Firstly, the gray image can be obtained by using template to smooth binary vein image, which overcomes defect that edge of binary image lack of transition points. Selecting a template and rotating it in different direction. We can get standards matching template with fusing and dilating vein rotated. Computing matching between image and its template judges whether vein is match or not. Experimental results show that the introduced method has higher recognition rate. Its recognition accuracy achieve 99.82%.(5) A match method using the modified Hausdorff distance algorithm based on the vein minutiae patterns(bifurcation points and ending points)is proposed. Since the vein pattern is represented as a set of two-dimensional points, the modified Hausdorff distance algorithm is proposed to evaluate the discriminating between all possible relative positions of the shape of vein patterns and achive to match. Experimental results show the algorithm is effective.(6) In this paper, a new technique called two-directional two-dimensional Kernel-based principal component analysisis analyzed. The feature is extracting by working simultaneously in the vertical and the horizontal directions of Finger-Vein images. Experimental results show the introduced method has the advantage of better running time and higher recognition rate, and the proposed method is t effective.