The Research Of Key Algorithm In Automatic Fingerprint Identification System

Fingerprint identification has been one of the most effective methods on personal identification from long time ago due to fingerprints' uniqueness and invariance. Automatic Fingerprint Identification System (AFIS) research is an important subject in the field of image processing and pattern recognition, and has high values and broad markets in future.This paper mainly focuses on the development of the key algorithms used in AFIS, including fingerprint image enhancement, feature extraction and matching identification. Experiments showed that the algorithms are effective.Author's work mainly includs:1) In the procedure of preprocessing, a fast fingerprints image enhancement algorithm applies Gabor filter to an image based on the estimated local ridge orientation and frequency and can adaptively improve the clarity of ridge furrow structures of input fingerprint images.2) To deal with the problem of unsmooth edges obtained by local binarization, an approach is proposed that dajing binaraization is used to get the threshhold of each local image block, and then global interpolation is used to calculate a new threshold for each block.,so can avoid the problem of unsmooth edges that brought by grouping.3) In terms of minutia feature extraction, an improved algorithm is applied to the fingerprint image which after thinning operation so that pseudo features can be deleted. Different algorithms are used to remove pseudo feature points caused byvarious reasons. Based on topology feature of ridge point and branch point, pseudo minutias are removed, and which is fundamental of fingerprint matching. 4) A matching algorithm is implemented in this paper based on minutia feature. By using ridge points and branch points as fingerprint feature, fingerprint matching fulfilled in polar coordinate space. Applying scalable delimitation box can solve nonlinear distortion of minutia matching in fingerprint image. Experiment shows that algorithms we proposed in this paper improve image processing and accelerates matching speed. The conclusion is valuable in both theoretical and practical.