Contactless fingerprint recognition using deep learning

Abstract

Contactless fingerprint recognition has become a popular field of research in biometrics in the last decade. Contactless fingerprint systems provide the advantages such as easy capturing and cost-effectiveness, along with the solutions to the problems in respect of hygiene, forgery, and latent fingerprint. Although many advancements have been taken place in this area, fingerprint recognition in contactless environment is still a challenging problem due to various constraints. For example, the presence of limited information in the image, background noise and low contrast between the ridges and the valleys. The contactless fingerprint system proposed in this work uses a Siamese model in deep learning framework to extract global features from a contactless fingerprint image. The method achieves an equal error rate (EER) of 10.07%, which is better than the EER obtained by the methods that employ handcrafted features namely, the minutia based NBIS Matcher and texture feature based Gabor filter bank. The score-level fusion of Siamese model, NBIS Matcher and CompCode yields the best matching performance with an EER of 3.53% on the contactless fingerprint dataset of HKPU Contactless 2D to Contact-based 2D Fingerprint Images Database Version 1.0.