Touchless Fingerprint Recognition with Capsule Networks and PCA Filtration Using Dual-Cross Generative Adversarial Networks
Khushboo Agarwal, Manish Dixit · Traitement du signal · 2024
Touchless fingerprint recognition is becoming increasingly popular as biometric authentication in terms of both ease and cleanliness.Furthermore, they offer advantages in terms of speed, robustness, and flexibility in challenging circumstances while also meeting the increasing need for touchless technologies in a post-COVID-19 era.These touchless fingerprint images have a unique quality that sets them apart from traditional ink-based and live-scan fingerprints.Existing touch-based fingerprint matchers often struggle to extract reliable minutiae features due to differences in contrast, illumination, and magnification.In contrast to touch-based systems, which have their own set of problems, such as the existence of latent fingerprints or deformation brought about by pressing fingers over a sensor surface, touchless acquisition processes have none of these problems.In this paper, a novel Dual-Cross Generative Adversarial Networks framework with Capsule Networks-based PCA filtration is proposed to accurately recognize the touchless fingerprint.In the proposed model, Capsule network-based PCA filtration is utilized for fast feature embedding with a convolutional architecture to collect spatial information.To handle all the diversification, Dual-Cross Generative Adversarial Networks is modeled to restore and recognize the fingerprint.The performance of the proposed system is assessed using two widely recognized datasets (the PolyU Cross dataset and the Benchmark 2D/3D dataset).The experimental results show that the proposed system achieves an accuracy of 99.51% and 99.13%, respectively, and significantly reduces the Equal Error Rate compared to the baseline.