End to End Fingerprint Verification Based on Convolutional Neural Network

Abstract

Fingerprint recognition has become one of the most reliable ways for human identification due to its uniqueness and consistency. The fingerprint matching problem is formulated as a classification system in which a model is learned to classify every two fingerprints as a genuine or impostor pair. Traditional approaches perform a feature extraction step before matching a fingerprint pair. On the other hand, recently convolutional neural networks (CNNs) have presented exceptional success for many image processing tasks such as face recognition. However, there have been only a few attempts to develop fully CNN methods to deal with challenges in fingerprint recognition problem. In this paper, a CNN-based fingerprint matching method has been developed. A key contribution of the proposed method is to directly learn fingerprint patterns from raw pixels of images. In order to achieve robustness and characterize the similarities comprehensively, incomplete and partial fingerprint pairs were taken into account to extract complementary features. Also, we proposed an end to end CNN approach that contains the feature extraction part of the trained AlexNet network. The network reached an EER of 17.5% on the FVC2002 dataset, that shows better results in comparison to the MinutiaSC and A-KAZE methods.