Improved AFIS for Color and Gray Image based on Biometric Triangulation

Abstract

This research presents a fingerprint image processing algorithm for personal automatic identification, which has been in development since 1998. It is principally based on the comparison of the fingerprint's biometric pattern between the fingerprint captured (original) in each session and the one stored in database. It is preferable to capture the image in color. The biometric pattern is formed by the Euclidean distances based on the triangulation of only three minutiae. This methodology locates the position and the type of each minutia to perform the triangulation. The applied metric is the statistic similarity obtained by the comparison of both biometric patterns. This technique enables one to solve translation and rotation problems. An original colored fingerprint is used in order to obtain more information about the fingerprint situation. The space color used is HCL, because it helps get a good skin color for an encrypt key, which is formed by each channel (HCL) in accordance with the skin color. This system has several applications due to its low cost and efficiency. Finally, the results obtained with this methodology were satisfactory since in all the experimental tests the system offered a rate of global success of 99 %.