Research Of Fingerprint Verification

As an identity verification method, fingerprint verification is one of the most important sub fields of biometrics. In real application of fingerprint verification, fingerprint images are compared on the chose processor based on the requirement in practice when they are captured by fingerprint sensor. The result of verification is showed in the terminal. This paper, based on the study of many fingerprint verification technology on computer, propose effective methods of image enhancement, deleting false minutiae and feature matching, which get a better experimental result. The algorithm is consist of normalize, computing orientation map, image enhancement, thinning, extracting minutiae and minutiae matching. Orientation map is expressed in two ways. One is successive, another is discrete. Image enhancement is the most importance step of minutiae extraction process. Gabor filter is usually an effect method for image enhancement. Because there is some noise in the low resolution fingerprint image, Gabor filters usually either remove the noise incompletely or enhance the image by mistake. The algorithm proposed in article contains not only linear but also nonlinear enhancement algorithm. Experimental results show that the enhancement algorithm improves accuracy of minutia extraction. The paper adopt the way of minutia extraction after two-pass thinning algorithm. Fasle minutiae are wiped off from minutiae. In minutiae matching procedure, we propose a new fingerprint feature named Neighborhood Direction Feature (NDF), which is consist of four numbers of crossing fingerprint ridge and relative orientations of fingerprint ridge. NDF is unconcerned with translation and rotation. Our experimental results shows that NDF bring down false accept and speed up matching procedure.