FINGERPRINT MATCHING USING MINUTIAE QUADRUPLETS

Abstract

Fingerprint matching faces several challenges resulting from the varying quality of fingerprint scanners, the weakness of some scanners in detecting fake fingerprints and the poor performance of fingerprint matching algorithms caused by the high intraclass variations between fingerprints of the same subject. The major contributions to the high intraclass variations include high displacement between an enrolled fingerprint and an input fingerprint of the same person; rotations and occlusions caused by non-uniform finger pressure on the scanner, differences in sensed fingerprint area, dry-skin or accidents; non-linear distortion as a 3 dimensional fingerprint image is sensed by a 2 dimensional fingerprint scanner. This paper proposes a new method of matching fingerprints using geometric features termed minutiae quadruplets. The performance of the properties of minutiae quadruplets were evaluated on unprocessed datasets of database_a (110 pairs of fingerprint images) of the Fingerprint Verification Competition (FVC) 2000 database.  The evaluation resulted in genuine matches of 89 out of 110 fingerprints; and a receiver operating characteristics of about 20% off the zero axes. This entails that the properties of minutiae quadruplets may not be beneficial for fingerprint matching of unprocessed fingerprints but may be beneficial for matching of processed fingerprints.