Fingerprint Matching - Hard Cases

Abstract: Today fingerprint matching software is widely used in applications such as mobile phones. Software for enrollment and verification of fingerprints is usually designed to work for "good" fingerprints, meaning that the fingerprint matching algorithms use minutiae locations. When such information is scarce due to scars, blisters or other damages, the algorithms do not work very well. In this work we demonstrate that gray scale matchers based on interest point detection and extracted local information can be used for matching fingerprints in such cases. Matchers with eigenvalue corner detection such as the Harris- and Shi and Tomasi- corner detectors has resulted in a matching performance of 5.92% false reject rate (FRR), which is a 0.97% improvement on the given database. Combining a gray scale matcher with an ordinary matcher can further reduce the FRR down to 2.54% and suggests that combining algorithms will result in a more secure system.