230 likes | 251 Views
Explore the history and implementation of fingerprint recognition using Gabor filter-based pattern matching. Learn about minutiae matching, reference point detection, and feature map creation in this detailed study. Understand how matching scores and biometric benchmarks like FAR and FRR are used for authentication. Test results show an EER of 1.88%.
E N D
Fingerprint Recognition by Matching of Gabor filter-based Patterns Diplomarbeit Aufgabensteller: Prof. Dr. Bernd Radig Betreuer: Dipl. Inf. Matthias Wimmer
Biometrics Idea: Authentification of human beings using physical characteristics History of the use of fingerprints: 19th century: Uniqueness of fingerprints 1998: FBI - IAFIS, Integrated Automatic Fingerprint Identification System Technische Universität München Markus Huppmann
Authentification (Workflow) • Enrollment • Detection of unique attributes • Creation of the template • Matching: Comparison of the template with other templates → Matching score → Decision: Acceptance or rejection (threshold) Technische Universität München Markus Huppmann
Minutiae Matching (1) Fingerprint recognition using ridge singularities: - Ridge bifurcation - Ridge ending Technische Universität München Markus Huppmann
Minutiae Matching (2) Technische Universität München Markus Huppmann
Minutiae Matching (3) Matching: → Matchingscore Technische Universität München Markus Huppmann
Problems • Fingerprints of dry or wet fingers • Non-overlapping areas → Global approach: Pattern Matching Technische Universität München Markus Huppmann
Pattern Matching Gabor filter-based Pattern Matching • Normalization • Segmentation • Reference point detection • Gabor filter • Creation of the Feature Map • Matching Technische Universität München Markus Huppmann
Reference Point Detection • Reference point defined as the point, where the ridges possess the highest curvature • Orientation map Technische Universität München Markus Huppmann
Gabor Filter (1) Sinusoid multiplied by a Gaussian function Technische Universität München Markus Huppmann
Gabor Filter (2) Gabor filter in direction 0° Technische Universität München Markus Huppmann
Gabor Filter (3) Technische Universität München Markus Huppmann
Creation of the Feature Map Tessellation → Template Technische Universität München Markus Huppmann
Creation of the Feature Map Technische Universität München Markus Huppmann
Matching (1) Comparison of the feature maps: Similar feature maps → low distance → "good" matching score → acceptance Technische Universität München Markus Huppmann
Matching (2) Technische Universität München Markus Huppmann
Matching (3) Different feature maps → high distance → "bad" matching score → rejection Technische Universität München Markus Huppmann
Matching (4) Technische Universität München Markus Huppmann
Tests • Database of 80 fingers with 4 fingerprints per finger • 2 Tests: • Genuine test: Matching of every fingerprint of the same finger (1A:1B, 1A:1C, 1A:1D, 1B:1C, … , 1C:1D) → "good" matching scores • Imposter test: Matching of the first fingerprint of every set with the first fingerprint of the other sets (1A:2A, 1A:3A, … , 79A:80A) → "bad" matching scores Technische Universität München Markus Huppmann
Biometric benchmarks • FAR: false acceptance rate • FRR: false rejection rate • EER: equal error rate optimal threshold where FAR = FRR Technische Universität München Markus Huppmann
Test results equal error rate = 1.88 % Technische Universität München Markus Huppmann
Questions? Technische Universität München Markus Huppmann