Automatic fingerprint interpretation and classification via contextual analysis and topological coding
William J. Hankley · OhioLink ETD Center (Ohio Library and Information Network) · 1968
A scheme for automatic interpretation and classification of single fingerprints is presented.Contextual analysis and topo logical coding by digital computer form the basis for design of the proposed system.Sets of fingerprints are commonly classified based upon the gross topological structure of the individual prints as loop, arch, whorl, etc.Individual prints are distinguished according to the topological pattern of certain fine features such as the core, delta, and minutiae which include ridge endings and bifurcations (forks).In the Battley system individual prints are partially classified according to a very small number of features presented in a pictorial classification guide.However, the number of classes by this method is limited.There is an urgent need for a clas sification system ^diich orders single prints.It might naturally be expected to be based upon the topological structure of the fine featm^es.The proposed topological coding scheme satisfies these properties, The most important operation in automatic fingerprint clas sification is the interpretive extraction of features.This is discussed explicitly.The proposed scheme "learns" the core location and the pattern of features in the sense that interpretation decisions are based upon previous interpretation experiences.This is accomplished by evolving a multilayer interpreted image of the fingerprints The scheme performs contextual analysis in order to cope with latent fingerprints and print noise due to smear and distortion caused by rolling, poor inking, and skin imperfections.The topological coding scheme analyzes each fingerprint and encodes the pattern of features into a string of code symbols.These symbols constitute code words and sentences which bear the print topological information, A natural ordering of code words induces an ordering of fingerprint patterns which is the basis for classification, A first generation ccmputer program is presented idiich danonstrates the feasibility of the proposed scheme.CHAPTER I