have been largely proprietary and vendor specific (i.e., different for each type of system). In fact, experienced latent print examiners have found that different systems will retrieve different stored prints in response to a given input map of features, and they have learned system-specific ways of annotating features on a latent print in order to maximize the success of each system’s (inferred) search algorithms. However, achieving broad-based AFIS interoperability will require baseline standards for these algorithms, so that fingerprint examiners can be assured of consistent feature mapping across systems. As mentioned previously, fingerprint examiners have learned by experience to provide different inputs to different vendors’ systems, often purposely leaving out information—knowing that the added input will degrade the search quality:

The examiner does not necessarily encode every point he can find in the latent print. LPU [latent print unit] examiners have learned through experience with the IAFIS program which types of points are most likely to yield a correct match. LPU Unit Chief Meagher told the OIG [Office of Inspector General] that examiners are taught to avoid encoding points in areas of high curvature ridge flow, such as the extreme core of a print. Unit Chief Wieners and Supervisor Green told the OIG that IAFIS does not do well when asked to search prints in which points have been encoded in two or more clusters separated by a gap. One reason is that IAFIS gives significant weight to the ridge count between points. If the ridge count between two clusters of points in a latent is unclear, IAFIS may fail to retrieve the true source of the print. Thus, an examiner will not necessarily encode every point that can be seen in a latent fingerprint, but rather may limit his encoding to points in a defined area in which the ridge count between points is clear.5

The fact that today’s systems often do not effectively utilize most of the available feature information and require substantial input from fingerprint examiners suggests that there is significant room for improvement. An ideal, comprehensive AFIS, for example, would be capable of automated:

  • reading of latent prints;

  • encoding of most features of usable quality, including those features identified as Level 1 (fingerprint classes such as whorl, arch), Level 2 (minutiae), Level 3 (pores, cuts), and ridge paths, together with a provision for including other features that could be defined by the vendor/user;


Office of the Inspector General, Oversight and Review Division, U.S. Department of Justice. 2006. A Review of the FBI’s Handling of the Brandon Mayfield Case, p. 119.

The National Academies of Sciences, Engineering, and Medicine
500 Fifth St. N.W. | Washington, D.C. 20001

Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement