Characteristic and necessary minutiae in fingerprints
by Johannes Wieditz
Date of Examination:2021-08-16
Date of issue:2021-10-22
Advisor:Prof. Dr. Stephan F. Huckemann
Referee:Prof. Dr. Stephan F. Huckemann
Referee:Prof. Dr. Dominic Schuhmacher
Files in this item
Name:Dissertation_Wieditz.pdf
Size:31.1Mb
Format:PDF
Abstract
English
Fingerprints feature a ridge line pattern inducing an undirected orientation field (OF) which usually features some singularities. Ridges vary in width, inducing a moderately varying ridge frequency (RF). In fingerprint recognition, a fingerprint is usually reduced to a point pattern consisting of minutiae, i.e. points where the ridge lines end or fork. Geometrically, minutiae can occur due to diverging ridge lines with a nearly constant RF or by widening of parallel ridges making space for new ridge lines originating at minutiae (and, indeed, combinations of both). We call these the geometrically necessary minutiae. In this thesis, we provide a mathematical framework based on vector fields in which orientation fields, ridge frequency as well as the number of geometrically necessary minutiae become tangible and easily computable using the provided algorithms and software. It turns out that fingerprints feature additional minutiae which occur at rather arbitrary locations. We call these the random minutiae, or, since they may convey fingerprint individuality beyond OF and RF, the characteristic minutiae. In consequence, a minutiae point pattern is assumed to be a realization of the superposition of two stochastic point processes: a Strauss point process (whose activity function is given by the divergence field) with an additional hard core, and a homogeneous Poisson point process, modelling the necessary and the characteristic minutiae, respectively. Given a minutiae pattern we strive for a method allowing for separation of minutiae and inference for the model parameters and consider the problem from two view points. From a frequentist point of view we first solely aim on estimating the model parameters (without separating the processes). To this end, we lay the foundations for parametric inference by deriving the density of the superimposed process and provide an identifiability result. We propose an approach for the computation of a maximum pseudolikelihood estimator and highlight benefits and drawbacks of this estimator on real and simulated data. Following a Bayesian approach we propose an MCMC-based minutiae separating algorithm (MiSeal) which allows for estimation of the underlying model parameters as well as of the posterior probabilities of minutiae being characteristic. In a proof of concept, we provide evidence that for two different prints with similar OF and RF the characteristic minutiae convey fingerprint individuality.
Keywords: Bayesian inference; biometrics; classification; divergence; Markov Chain Monte Carlo; parameter estimation; spatial point processes; maximum pseudolikelihood estimation