Automatic personal identification using fingerprints
Anil Kumar Jain, Hong Lin · 1998
An accurate automatic personal is critical in a wide range of application domains such as national ID card, electronic commerce, and automated banking. Biometrics, which refers to automatic of a person based on her physiological or behavioral characteristics, is inherently more reliable and more capable in differentiating between an authorized person and a fraudulent imposter than traditional methods such as passwords and PIN numbers. Automatic fingerprint is one of the most reliable biometric technology. In this thesis, our objective is to design a fingerprint-based biometric system which is capable of achieving a fully automatic positive personal identification with a high level of confidence. We have identified and explored the following issues: (i) feature extraction finding representative features from an input image for the purpose of fingerprint matching, (ii) image enhancement--improving the clarity of ridge structures of fingerprint images to facilitate automatic extraction of features or for visual inspection, (iii) minutiae matching--determining whether two sets of features (minutiae patterns) are extracted from the same finger, (iv) integration of multiple biometrics--improving the performance of a biometric system by combining several biometrics (e.g. fingerprint, face, speech, etc.), and (v) fingerprint classification--assigning a fingerprint into one of several pre-specified categories according to its pattern formation. We have designed two prototype biometric systems: (i) a verification system which uses only fingerprints to authenticate the identity claimed by a user, and (ii) an integrated system which combines face recognition and fingerprint verification to make a personal identification. Our systems have been evaluated extensively on a large number of fingerprint images captured with the traditional inked method and more recent inkless optical scanners. Experimental results show that our systems perform very well on these data sets.