Fingerprint Comparison. II: On the Development of a Single Fingerprint Filing and Searching System

Abstract

A FORTRAN program has been written to compare the minutiae coded in an “unknown” fingerprint with the minutiae of fingerprints stored in a data base. The criteria for matching are scores based on the type and number of minutiae matched. Minutiae of low relative frequency have higher scoring weights than those that are more common. The matching mechanism is tested by using a fingerprint coded several times, first by a single individual and then by six other individuals who have no previous knowledge about fingerprints. These tests yield satisfactory results.

Fingerprint features-statistical analysis and system performance estimates

“…For a random vector describing the outcome of the th cell, Sclove's probability of a set of Galton features is represented as (23) For simplicity, the four preceding adjacent cells are referred to as matrix , where the conditional probability of (23) is reduced to (24) Assuming that the cells in exert influence over cell , the probability of feature occurrence involves the number of adjacencies as well as the different types of features ; if a cell is occupied, there are still 12 different event possibilities (using data from Table 1 in Appendix III). Sclove's probability of a minutia (or Galton feature) occurrence in a cell is then (25) where is the number of empty cells with adjacencies, is the number of cells containing the probability of minutia type is the probability of an event's occurring, and is the number of occupied cells with adjacencies. Sclove found that as (the number of adjacencies) increases, the probability of the central cell's being occupied increases monotonically.…”

A Novel Global Pattern Recognition Algorithm

“…Early fingerprint identification algorithms were minutiae based methods developed for the FBI's AFIS *31]. Early algorithms divided fingerprints according to a grid, and noted which grid sections contained minutiae [32]. Each grid location was given a value based on its minutiae type [32].…”

“…Early algorithms divided fingerprints according to a grid, and noted which grid sections contained minutiae [32]. Each grid location was given a value based on its minutiae type [32]. Grid values from different fingerprints could then be compared for matching minutiae.…”

“…Grid values from different fingerprints could then be compared for matching minutiae. [32]. The benefit of this method was that it closely resembles the process followed by a human fingerprint examiner.…”

“…The benefit of this method was that it closely resembles the process followed by a human fingerprint examiner. The disadvantage to this algorithm was that fingerprints must be correctly aligned (offset and rotation) according to a strict set of rules [32] The inherent variability and elasticity of fingerprints makes alignment a difficult task [18]. Not all fingerprints share the same global patterns or common minutiae, so it is difficult to develop a set of rules that will always align a fingerprint in the desired way.…”