Evaluation of Contactless Versus Contact Fingerprint Data, Phase 2 (Version 1.1)

• NCJ Number: 249552
• Author(s): Lars Ericson; Stephen Shine
• Date Published: May 2015
• Length: 117 pages
• Grant Number(s): 2010-IJ-CX-K024; 2014-ZD-CX-K001
• Sponsoring Agency: National Institute of Justice (NIJ)
• Corporate Author: ManTech Advanced Systems International, Inc.
• Sale Source: US Dept of Justice NIJ Pub
• Publication Format: Document; Document (Online)
• Publication Type: Test/Measurement; Research (Applied/Empirical); Report (Technical); Report (Study/Research); Report (Grant Sponsored)
• Language: English
• Country: United States of America

Annotation

This project compared fingerprint data collected from traditional contact-based legacy fingerprint (LFP) devices with fingerprint data produced by next-generation contactless fingerprint (CFP) scanners.

Abstract

This proved to be the second quantitative demonstration by a third party that fingerprints collected under ideal conditions from LFP and CFP devices can be matched against each other in a statistically meaningful way. This report concludes that the experimental methodology used (data collection and analysis) can be used to determine a comparative match performance among LFP and CFP using two-dimensional (2D) projections. In addition, researchers conclude that the CFPv2 dataset has sufficient breadth and depth for investigating the performance and interoperability of CFP and LFP devices. There are few biometric dataset resources that cover as many devices that capture the same subject pool. The Morpho Trak Finger-On-The Fly (FOTF) and the ANDI On-The-Go (OTG) contactless fingerprint devices performed well against a LFP. The contact mobile devices performed well (both plain and rolled printed). The results were comparable to fixed livescan systems. The matching accuracy of a probe set was highest when matched against its same type; and the OTG Binary images led to a higher match accuracy than OTG Grayscale images. In addition, the NIST Fingerprint Image Quality (NFIQ) score distributions of some CFP dataset were not fully consistent with the resulting match accuracy. Researchers conclude that the NFIQ algorithm may not be optimized for predicting match performance for certain kinds of CFP images. Researchers also conclude that some CFP capture devices that perform poorly against third-party databases may have a role in low-security verification applications. They suggest that test parameters should mirror operational end-use rather than relying solely on amplified generic test procedures. 16 figures, 10 tables, and appended collection device summaries