U.S. flag

An official website of the United States government, Department of Justice.

NCJRS Virtual Library

The Virtual Library houses over 235,000 criminal justice resources, including all known OJP works.
Click here to search the NCJRS Virtual Library

MEASURING AFIS MATCHER ACCURACY

NCJ Number
148185
Journal
Police Chief Volume: 61 Issue: 4 Dated: (April 1994) Pages: 147-151
Author(s)
M K Sparrow
Date Published
1994
Length
5 pages
Annotation
The author discusses automated fingerprint identification systems (AFIS) for the sake of potential buyers.
Abstract
Buyers--cities, States, and nations--typically make AFIS decisions based on political, sole-source, or pragmatic concerns, rather than an informed understanding of system performance. The most crucial component of AFIS, accuracy, is also the most difficult to measure. Other factors such as price, space requirements, user friendliness, and computability are much easier to assess. Buyers should not settle for "99.9-percent accuracy." It is important to have a contextual interpretation of other users' claims of satisfaction with their systems. For one thing, other users have an interest in defending their past buying decisions. This and other important considerations are discussed: ten- print tests using the same cards twice; generating artificial latent images from rolled prints; overestimating the ten-print comparison problem; using "dabs" as latents; advance submission of test sets; failure to correct for size of database; use of score-dependent measures; use of multiple thresholds; latent test set based upon previous performance; minimum total error; mismatch score distribution percentiles; and ranked lists. Figure, 3 endnotes