美国数学建模竞赛特等奖论文——2004 A O Not Such a Small Whorl After All【数学建模】.pdfVIP

Not Such a Small Whorl 245 Not Such a Small Whorl After All Brian Camley Pascal Getreuer Bradley Klingenberg University of Colorado at Boulder Boulder, CO Advisor: Anne M. Dougherty Summary Fingerprint identification depends on the assumption that a persons fin- gerprints are unique. We assess the truth of this assumption by calculating the total number of distinct fingerprints. We assume accurate fingerprints (ignoring procedural error) that are de- fined by 12 points of detail or minutiae. The number of distinct fingerprints depends also on the number of potential positions of these minutiae. Two his- torical methods and a geometric analysis estimate there to be 1,400 positions, a figure confirmed by our algorithm for counting ridges in a fingerprint. We create two models to estimate the number of unique fingerprints: * One model computes fingerprints as arrangements in minutiae; * the other extrapolates the number of fingerprints from the Shannon entropy of the information that defines a fingerprint. These two models agree to within an order of magnitude that there are 5 x 1033 unique fingerprints, a compelling validation of our general approach. To handle the large number of fingerprints, we implement an approximation for the calculation of probabilities. Given a cumulative world population of 120 billion [Catton 2000], the probability of two people ever having the same fingerprint is 1.4 x 10-6. The probability of two humans living today sharing a fingerprint is 3.5 x 10-15, which suggests that fingerprints are a theoretically more reliable method of identification than DNA analysis, which has a false positive probability of 10-9. None of these calculations take into account procedural