Instructions

U.are.U SDK - Developer Guide 50
The .NET API
Creating FMDs from images
The CreateFmdFromFid() method can be used in two ways:
1. Extract fingerprint minutiae from a raw image and create an FMD.
2. Extract fingerprint minutiae from an FID and create an FMD.
The following limitations are applied to the raw images and FIDs:
• 8 bits per pixel
• no padding
• square pixels (horizontal and vertical dpi are the same)
The size of the resulting FMD will vary depending on the minutiae in a specific fingerprint.
Comparing Fingerprints
Identify() identifies a single FMD against an array of FMDs. This function takes as inputs:
• A single view in an FMD
• An array of FMDs (each FMD can contain up to 16 views) to compare
• The desired number of candidates to return
• The threshold for False Positive Identification Rate (FPIR) that is permitted
and returns matches as an array of integer pairs which provide the finger index and view index of each match.
Each time a view has a score lower than the threshold FPIR, that view is marked as a possible candidate. Then when all
possible candidates are identified (i.e., they meet the threshold), they are ranked by their score. Finally, the function
returns as many candidates as requested, based on the candidates with the lowest dissimilarity score. For a discussion
of setting the threshold as well as the statistical validity of the dissimilarity score and error rates, consult
NIST
Fingerprint Image Quality (NFIQ) on page 14
.
Compare() takes two single views from two FMDs and returns a dissimilarity score indicating the quality of the
match.
The majority of applications should use the
Identify method to implement both identification and verification.
However, in a few special cases, e.g., using multi-modal biometrics, or doing statistical risk assessment, the
Compare
method allows you to compare two FMVs to determine their actual degree of dissimilarity. This is useful for accuracy
testing and diagnostics and is not intended to be used in final applications for actual fingerprint recognition. The
Compare method returns a dissimilarity score with values:
• 0 = fingerprints are NOT dissimilar (i.e., they MATCH perfectly).
•
maxint (#7FFFFFFF or 2147483647) = fingerprints are completely dissimilar (i.e., DO NOT match).
• Values close to 0 indicate very close matches, values closer to
maxint indicate very poor matches.
The table below shows the relationship between the scores returned from
Compare and the false match error rates
observed in our test. The dissimilarity score distribution is estimated based on our internal testing, and may not be
representative of the actual rate that will be observed in deployment.
Dissimilarity Score False Match Rate
2147483 .1%
214748 .01%
21474 .001%
2147 .0001%