Page 111 - Automated Fingerprint Identification Systems (AFIS)

96 AUTOMATED FINGERPRINT IDENTIFICATION SYSTEMS

of 500ppi for image capture, many agencies are opting to capture these images
at higher resolutions, e.g., 1,000ppi. Although these higher resolution images
require four times as much storage and bandwidth for transmission, they
provide much better clarity and ridge detail. Therefore, such AFIS configura-
tions, made possible by the rapid advancements in the field of identification,
the decrease in computer costs, improved system reliability, and a massive
amount of state and federal funding, naturally have wide support among the
members of the latent print community.

5.1.5 MATCHERS

In the configuration shown in Figure 5.1, the matchers, which house the
extracted image characteristics, are located on the right and bottom. There are
multiple sets of matchers for various types of searches. When a search is initi-
ated, the extracted image characteristics of the search record are compared or
searched against the extracted image characteristics already on file. Scores are
assigned to the candidates produced by this search that indicate the relation-
ship of the image characteristics on file to the image characteristics of the
search print. If 75 minutiae of a search image exactly match 75 minutiae of an
enrolled record in the matchers, the enrolled record would be given a very high
score. It might have more than 75 minutiae, but the large number of match-
ing minutiae creates an extremely high probability that the two images are from
the same person.
As shown in Figure 5.1, there are three major components to the matcher
subsystem: the matcher controllers, the string controllers, and the matchers
themselves. Each has a specific function in the identification process.
The matcher controller is the computer that controls the operation of that
matcher subsystem. Transactions from the AFIS system and responses back to
it are channeled through the matcher controller.
Matchers may be arranged in an array, or string. The string of matchers con-
tains all the image characteristics for the particular record type. For example,
TPid matchers contain the image characteristics for all the tenprint records,
that is, the two index fingers (or possibly two thumbs, or both) of everyone who
has a tenprint record on file. These matchers are commercial-off-the-shelf
(COTS) computers that reach a storage capacity on an estimated fill rate. When
the matcher fills to a desired percentage, it is sealed from the addition of new
records and a new matcher is added to the string. Many systems support two or
more redundant matcher strings, which allow two searches to be conducted
simultaneously, one searching through string 0 and the other through string 1.
This redundancy allows the systems to meet throughput requirements for