Extended Feature Set and Touchless Imaging For Fingerprint Matching

Extended Feature Set and Touchless Imaging For Fingerprint Matching By Yi Chen In recent decades, fingerprint matching has undertaken a tremendous transition from a tedious manual procedure for criminal investigation to the most widely deployed biometric technology for government and civilians applications. In this thesis, we address two critical issues related to this transition: i) automatic systems have not fully utilized the knowledge gained by forensic experts in manual fingerprint matching (e.g., extended features, matching with distortion); ii) interoperability between advanced sensing technology and legacy fingerprint databases has not been fully achieved in automatic systems. To address the first issue, we investigate the use of extended features, often utilized in manual fingerprint matching, in automatic systems. These extended features include ridge skeletons, pores, dots and incipients. We propose methods to automatically extract and compare these extended features in a hierarchical fashion. Our experiments show performance improvement from each of the proposed extended features in live-scan matching on MSU database (full vs. full and partial vs. full). We also show that ridge skeletons are more effective than pores, dots and incipients in improving latent matching on NIST-27 database (latent vs. roll). In addition, we conduct statistical analysis on the individuality of fingerprints using extended features, demonstrating their discriminative nature both theoretically and empirically. To address the second issue, we investigate the interoperability between a new fingerprint sensing technology based on touchless imaging and the legacy rolled fingerprint data. We propose a non-parametric virtual rolling method to unwrap the 3D touchless fingerprints into 2D rolled-equivalent fingerprints. We also develop a quality measure and an enhancement algorithm for the unwrapped touchless fingerprints. Our experiments on the TBS database demonstrate effectiveness of the proposed methods in achieving compatibility in matching touchless fingerprints with legacy rolled fingerprints.