Design of a Finger Vein Recognition Algorithm Based on Improved ResNet

This paper presents an improved ResNet-18-based algorithm for finger vein recognition to address the challenges posed by insufficient data and difficult recognition tasks. By leveraging a pre-trained ResNet-18 model from the large-scale ImageNet dataset, the model benefits from the pre-learned complex features and patterns. Transfer learning was applied to fine-tune the entire model on a 106-class finger vein image dataset, optimizing both runtime and recognition accuracy. The experimental results demonstrate that the improved model achieves a recognition accuracy of 97.525%, outperforming the original ResNet-18 by 2.9%. This improvement highlights the model’s effectiveness in finger vein image classification.