The Research On Algorithm Of Automatic Fingerprint Classification

Fingerprint identification technology, which is used widely in all kinds of identification and identify fields, is one of the most popular technologies of biometrics identification. There are many algorithm focused on the fingerprint identification technology to date, and a lot of commercial products with automatic fingerprint identification system and research results have been released. But these products and results still can't meet certain specific application requirements, so there are many aspects and improvement worth studying.Fingerprint classification system mainly includes these aspects such as, fingerprint collection, segmentation, feature extraction and classification. We will research and analyze above aspects one by one, and make corresponding improvement on these algorithms in this paper.(1) Fingerprint segmentation: In this paper, based on the conventional variance method, directional chart method and the maximum between-class variance method (OTSU), we introduce a new fingerprint segmentation method that combine the variance method and the maximum between-class variance method (OTSU). The experiment results demonstrate that the new method can give the more accurate segmentation of foreground and background. Moreover, the new segmentation method can save time and space resource on account that the new method do not need to adjust appropriate threshold according to the empirical value.(2) Fingerprint feature extraction: We propose an improved finger feature extraction based on the research the conventional Poincare index method. The improved fingerprint feature extraction algorithm can attain smoother fingerprint direction chart than the conventional method. It can acquire the more accurate singular point positioning, more powerful robustness of algorithm.(3)Fingerprint classification: The fingerprint classification introduced in this paper that possesses the simple algorithm, short calculating time, low resource consumption and easy feasibility in experiment can achieve practical requirement based on the fact that it can exactly extracts the fingerprint features.We realize all algorithms mentioned above by means of"Matlab". Experiments indicate that all new algorithms introduce in this paper greatly improve the segmentation, features extraction and classification of fingerprints.