 Cogent Results in NIST Evaluation
The Fingerprint Vendor Technology Evaluation (FpTVE) 2003 was conducted on behalf of the Justice Department and fulfills part of NIST's mandate under the Patriot Act to certify biometric technologies that may be used in the U.S. Visitor and Immigrant Status Indicator Technology program. The NIST study examined 34 commercially available systems used to match fingerprints and was the most extensive test of the technology ever conducted by a government agency. 48,105 sets of fingerprints from 25,309 people, with a total of 393,370 distinct fingerprint images, were used to enable thorough testing.
From the evaluations of the commercial systems provided by 18 companies, the NIST study concluded that Cogent provided one of the top three most accurate fingerprint algorithms. As one of the most accurate systems, Cogent has consistently very low error rates across a variety of large-scale system data sets.
Although FpVTE did not intend to measure the resources required to achieve FRR and FAR levels (such as equipment and response time), these factors will have a significant bearing on cost to develop, deploy, and operate a system for an end user. For the FpVTE benchmark, the resources needed for searching with the Cogent algorithms were significantly less than some other vendors within the group of most accurate systems. Also, Cogent conducted the large scale system test in eight days while it took the vendor in the first position nineteen days.
SlapSeg04 is being conducted by the National Institute of Standards and Technology (NIST) on behalf of the Department of Justice (DOJ) Justice Management Division (JMD), IDENT/IAFIS Integration Project. This evaluation was conducted to determine the accuracy of existing slap segmentation algorithms on a variety of operational-quality slap fingerprints. Slaps are also known as four-finger simultaneous plain impressions. Slap segmentation is the process by which a single image containing four fingerprint images is divided into four images of the individual fingers. Cogent's algorithm was one of the most accurate algorithms used to segment slap fingerprint images into individual fingerprint images.
Under these independent tests the Cogent algorithm achieved an average TAR of 99%.
 NISTIR 7221 "Studies of One-to-One Fingerprint Matching with Vendor SDK Matchers"
This test evaluated the accuracy of current SDK (Software Development Kit) based COTS (Commercial Off-The-Shelf) fingerprint matching systems for one-to-one verification applications. Fingerprint matching systems from twelve vendors were evaluated.
NISTIR 7249 "Two Finger Matching with Vendor SDK Matchers
The two finger matching evaluation is an extension of the above testing used to evaluate the accuracy that can be achieved by combining the index finger scores to achieve a match. These results are based on the SDK matchers provided for the original single finger SDK testing.
This report discusses the flat-to-flat matching performance of the US-VISIT fingerprint matching system. Both one-to-many matching used to detect duplicate visa enrollments and one-to-one matching used to verify the identity of the visa holder are discussed. With the proper selection of an operating point, the one-to-many accuracy for a two-finger comparison against a database of 6,000,000 subjects is 95% with a false match rate of 0.08%. Using two fingers, the one-to-one matching accuracy is 99.5% with a false accept rate of 0.1%.
These results were achieved using a Software Development Kit (SDK) supplied by Cogent that is the same algorithm used in US-VISIT. NIST study concluded that "The Cogent image quality is a good predictor of the IDENT one-to-many matching performance. The best quality images (quality 1), produce a TAR of 99% at a FAR of 1%. The worst quality images (quality 8), produce a TAR of 53% at a FAR of 1%."
Moreover, "Cogent image quality is a good rank statistic for all the algorithms tested for all the datasets used. The error rate of the best (quality 1) fingerprints is always lower than the error rate of any other image quality level, and the error rate of the worst (quality 8) fingerprints is always the highest..." |
