Minutiae Detection: An Image Exploring Agent-Based Model

Abstract

Identification through fingerprints is one of the most precise methods to determine a person’s identity. For years, this task has been carried out manually, but, due to the great number of fingerprints involved in the comparison, the implementation of an Automated Fingerprint Identification System (A.F.I.S.) emerged as a necessity. A critical process in automated fingerprint identification is minutiae extraction, process in which the presence of image noise or filth will generate undesired spikes and breaks that will lead to a considerable number of spurious minutiae. To overcome this problem, before minutiae detection, a usual process applied is a post-processing of the thinned image or pruning step, being the latter not always successful. In this paper, we introduce a new minutiae detection method, without the need of a pruning step, based on image exploring agents. First, reactive agents are used over the thinned fingerprint image, that detect locations of interest that could be minutiae using an efficient coefficient, which we propose, and to our knowledge, has not been applied before for minutiae detection. Then, several agents run through the image, starting at the locations of interest, to determine whether they are real minutiae or not. Having a more reliable method for minutiae detection would lead into a better performance of the whole recognition system.