Matching Fingerprints with a Toroidal Iterative Closest Point Algorithm

Pitcher, Courtney and Marais, Patrick and Darlow, Luke (2020) Matching Fingerprints with a Toroidal Iterative Closest Point Algorithm, Proceedings of Conference of the South African Institute of Computer Scientists and Information Technologists, 14-16 September, Cape Town, 51-57, ACM.

[thumbnail of FingerprintMatching.pdf] Text
FingerprintMatching.pdf - Accepted Version

Download (1MB)


We develop a generalisable Cartesian-toroidal Iterative Closest Points (ICP) fingerprint matcher. Both 2D and 3D minutiae are cast as point clouds that we augment with additional features. Iterative Closest Points (ICP) is immediately applicable to these data structures, yet nonetheless relatively unexplored in the fingerprinting domain. We apply our ICP-based method to conventional 2D minutiae and 3D features extracted from Optical Coherence Tomography (OCT) scans of fingertips. We show that ICP is a viable strategy to fingerprint matching using the diverse features in the internal fingerprint skin. Using 3D minutiae alone gave an Area Under the Curve (AUC) of 0.961, and 3D minutiae augmented with mean local OCT intensity gave an AUC of 0.973. Regarding 2D minutiae, our method offers a significant improvement over the baseline NIST Bozorth3 algorithm: an AUC of 0.94 versus 0.86 on an artificial dataset generated with SFinGe. In addition, ICP incurs only nominal computation cost when additional features are added.

Item Type: Conference paper
Subjects: Security and privacy
Security and privacy > Security services > Access control
Security and privacy > Security services > Authentication > Biometrics
Alternate Locations:
Date Deposited: 03 Dec 2021 10:51
Last Modified: 03 Dec 2021 10:51

Actions (login required)

View Item View Item