A series of fingerprint matching studies have been conducted on an experimental laboratory system called the Verification Test Bed (VTB). The VTB is a collection of commercial off the shelf (COTS) computer hardware, an open-source operating system, and a suite of public domain application software. Results are presented that compare various sources of fingerprints, assess the image quality of fingerprints by analyzing the matcher scores for inked and live-scan impressions, and study the trade-offs of matching rolled fingerprints with plain impressions. Database size in these studies range from 216 to ~600,000 people. Performance statistics are primarily reported for single-finger matching; however, results from two different approaches to two-finger fusion matching are also presented. At a false accept rate (FAR) of 1%, the best two-finger true accept rate (TAR) is 99% while the worst single-finger TAR is 71%. This report illustrates the wide range of image types and quality that exist in government fingerprint databases and the effect this variability has on the accuracy of matching using a single algorithm.
Citation: NIST Interagency/Internal Report (NISTIR) - 7020Report Number:
NIST Pub Series: NIST Interagency/Internal Report (NISTIR)
Pub Type: NIST Pubs
fingerprint, identification, matching, system evaluation, verification