FV-DGNN: A Distance-Based Graph Neural Network for Finger Vein Recognition
Jie Chang, Taotao Lai, Luokun Yang, Chang Fang, Zuoyong Li, Hamido Fujita · IEEE Transactions on Instrumentation and Measurement · 2023
As a promising biometric identification technology, finger vein recognition has gained considerable attention in the field of information security due to its inherent advantages such as living body recognition, non-contact operation, and high security. However, existing models often focus on pairwise matching of low-contrast infrared finger vein images, overlooking the underlying relationships among the matching information. To address this limitation, we propose a Graph Neural Network (GNN) model that captures the distance-based inter-relation between multiple pairs of samples. Specifically, we design an architecture to obtain a binary finger vein mask image, which guides the model to capture high-level features of finger vein regions while ignoring noises behind non-finger vein regions. Moreover, a distance-based GNN architecture, which models the distance distribution between multiple pairs of finger vein images by fusing the distance information propagated along edges, is proposed to determine the matching degree between each pair of images. Furthermore, to further expedite the proposed model in application, the depth-wise separable convolution layer is adopted in the encoder component of a Convolutional Neural Network (CNN) architecture to reduce the parameters significantly. Extensive experimental results on three public databases have verified the effectiveness of our proposed model.