Research And Implementation Of Ridge-Based Automatic Fingerprint Verification Algorithm

Due to the uniqueness and invariability of fingerprint, as well as the feasibility and practicability of fingerprint identification technology, identification based on fingerprint is becoming one of the most popular personal authentication technologies. Although many fingerprint products have been offered nowadays, the identification technique has not been perfected yet. So how to realize more reliable and accurate automated fingerprint identification systems (AFIS) is still a difficult research subject. Because the social development requires better system performance, the study in this field still possesses important theoretical significance and practicable value.In this paper, a whole set of fingerprint verification algorithm has been proposed, based on summarizing and assimilating the research production in the field of AFIS. This algorithm is mainly composed of two sections, fingerprint pre-processing and fingerprint identification.Fingerprint pre-processing includes image segmentation, orientation estimation, orientation smoothing, ridge frequency estimation, image normalization, image filtering, image binarization and thinning. The main purpose of image normalization is to set the contrast and grey value of original images to a fixed level, which can supply a uniform ground for further image manipulation. Ridge orientation and ridge frequency are two prominent characteristics of fingerprint image. In order to make the best of thorn, Gabor filter is applied to enhance the quality of fingerprint image and the resistance to low image quality of AFIS system. A kind of simple but effective thinning algorithm is also proposed in this paper, it assure strictly that thinned lines be in the centre of ridge lines, and the whole image be thinned thoroughly, without any burr and twist.Traditional fingerprint identification algorithms based on minutiae only use a small part of information in fingerprint image, that is the position and orientation of the minutiae, as character to match, while other abundant structural information contained in image are lost, thus they hardly adapt to the change of fingerprint, and their robust to noise are doubted. Based on former research, a new kind of fingerprint identification algorithm based on ridge line-matching is proposed in this paper. In this method, the input fingerprint, which has been pre-processed to be a thinned binary image with 1-pixel line width, is directly aligned to the registered fingerprint, which has been thinned and then extended to be a binary image with certain pixel width. It use ridge line pixel to match instead of ridgeend and bifurcation, thus there is no need to extract fingerprint minutiae or to carry out complex post-processing, while matching reliability is enhanced. This algorithm is realized by block orientation rough matching, ridge line fine matching in template and ridge matching in windows with two levels of judgement.Experimental results show that our system has a relatively high rate of fingerprint recognition and a strong ability resist image rotation and displacement. In addition, it is proved to be relatively high effective in recognizing damaged or blurry fingerprint images.