MINUTIAE-BASED FINGERPPRINT IDENTIFICATION AND VERIFICATION

• Main Author: Aditya Puspa Kania, Riska
• Format: Theses
• Language: Indonesia
• Subjects: Matematika
• Online Access: https://digilib.itb.ac.id/gdl/view/37088
• Institution: Institut Teknologi Bandung

Fingerprint is unique to everyone. There are no two fingerprint ridge patterns are ever exactly alike, compared one to other people. Hence, fingerprint is used widely for recognizing a personal identity, such as forensics purposes and security. Fingerprint can be recognized based on minutiae. A minutiae has coordinate, angle and type, truncation or bifurcation. The quality of fingerprint image influences the accuracy of the result of fingerprint recognition. Fingerprint image enhancement applied to reduce spurious minutiae during minutiae extraction. Limitations size of search of fingerprint on database reduce the time of searching. Multilayer Perceptron Artificial Neural Network (MLP ANN) classified fingerprints into classes during identification. The optimal structure of MLP ANN is consider by varying learning rate value and numbers of hidden layer units. The integrated MLP ANN - Particle Swarm Optimization (PSO) is also used to optimize MLP ANN structure. An offline character identification proposed by Märgner is used to be an input for MLP ANN. MLP ANN trained by using FVC2002 databases and gives accuracy below 50%. Re-alignment fingerprint image is suppose to be implemented to increase the accucary of MLP ANN. A fingerprint compared to all fingerprints on the same class as a graph, where vertices represent minutiaes and edges are connection between two minutiaes. Here, a graph matching problem need to be solved to verify that two fingerprints are equal, using graduated assignment method.