Advances in Fingerprint Technology.pdf

minutia orientation results in regions with converging or diverging ridges.
Thus, if the overall pattern of ridge flow is known, information is also available
concerning the orientations and distribution of minutiae. This aspect of
minutia distribution has not been included in any of the fingerprint models,
but it is obviously an essential element of any comprehensive treatment.
Orientation of Minutiae
Along the ridge flow, two fundamentally distinct orientations of minutiae
should be recognized. With the exception of the simple dot, minutiae are
formed when ridges are either added to or lost from the system. Minutia
orientations, similar to ridge counts, are robust to fingerprint distortions and
provide objective criteria for comparison.
Variation in Minutia Type
The difficulties encountered when using compound minutiae have been discussed.
Minutiae are best considered as one of three fundamental types: the
fork, the ending ridge, or dot. Compound varieties arise when the fundamental
types are positioned close to one another. Relative frequencies of the
fundamental types should be used, but correlations with the pattern of ridge
flow, neighboring minutia types, and minutia density need to be considered.
Variation among Prints from the Same Source
Treatment of variation among fingerprints from one finger is a crucial element
of a fingerprint comparison model. We have noted that ridge counts
and minutia orientations are robust to fingerprint distortions. Ridge spacing,
curvature, and distance between minutiae are variable. Some criteria must
be given for acceptable variation in these parameters. The relevant issue is
for a particular fingerprint of unknown origin, what is the set of known
fingerprint configurations that we would judge to be in agreement? We must
characterize this set to calculate the chances of encountering one of its members.
The problem is not simple. If the unknown fingerprint appears distorted,
we would naturally tolerate more variation. We would also accept
more variation as the distance between minutiae increases. A fingerprint
model must incorporate these features into the comparison criteria.
Connective ambiguities must also be allowed. As noted by Roxburgh, the
quality of the unknown print determines the magnitude of the allowance. In
the extreme, each minutia can be considered as one of three possibilities and
the ridge count can be affected. It is unrealistic to allow connective ambiguity
for all minutiae in the general case, but it is unremarkable to encounter a few
ambiguities — even in excellent prints. The amount of ambiguity to be tolerated
must be established, based on the quality of the unknown fingerprint, so that
the set of acceptable known configurations can be established.