Design of Artificial Neural Networks to Recognize Fingerprint Patterns

Abstract

Image or pattern recognition system is one of the branches in computer science, this system can help the processing of fingerprint patterns, especially in the banking, police and users of other institutions who really feel the importance of using fingerprints. Several stages in fingerprint pattern image recognition are through the process of scanning, then the resulting digital fingerprint image is converted to a certain value, among others, the threshold process, the division of images, and representation of input values. The training process is carried out using two treatments: the first with a different level of understanding and the second training with different unit numbers, the best training is obtained with a level of understanding of 0.3 and the number of hidden units 10 by producing a short training time and relatively small errors. Fingerprint pattern recognition is done by two trials, based on 1 number of training patterns and 5 number of training patterns. From the research data, the ability of the system to recognize output patterns is greater if the number of training patterns increases, with a number of 1 training patterns, the system is able to recognize 50% external patterns while the 5 system training patterns are able to recognize 70% output patterns.