Minutiae-Based Fingerprint Identification Using Gabor Wavelets and CNN Architecture

Abstract

Fingerprint identification is still a challenging issue for confident authentication. In this study, we present a methodology that comprises pre-processing, minutiae detection, and Gabor wavelet transform. Both Gabor wavelet and minutiae features, such as ridge bifurcation and ending enhancement, represent the significant information belonging to fingerprint images. Pre-processing algorithm affects minutiae extraction performance. So we use the dilation morphological operation and thinning for the enhancement. Then Gabor wavelet transform is applied to minutiae-extracted images to increase the identification performance. The classification problem is solved using a proper convolutional neural network (CNN) with a three-layer convolutional model and appropriate filter sizes. Experimental results demonstrate that the classification accuracy is 91.50% and the proposed approach can achieve good results even with poor quality images.