研究目的
To learn the features representative of a finger vein that is more discriminative and robust than handcrafted features and to address the issue of translation and rotation in vein imaging.
研究成果
The proposed method achieves excellent performance on several public datasets, demonstrating the robustness and superiority of our approach. The results of the validity evaluation experiment further demonstrate the effectiveness of our method.
研究不足
The finger vein area is unable to provide enough samples with translation and rotation, making it difficult for CNN to directly obtain the robust feature for translation and rotation.
1：Experimental Design and Method Selection:
A deep convolutional neural network (CNN) model, named the Finger-vein Network (FV-Net), is proposed.
2：Sample Selection and Data Sources:
Three public databases, including SDUMLA_HMT-FingerVein, FV-USM, and MMCBNU_6000-FingerVein, are used.
3：List of Experimental Equipment and Materials:
Not specified.
4：Experimental Procedures and Operational Workflow:
The FV-Net is trained with a classification task to learn an effective feature representation. A template-like matching strategy is proposed to match the finger-vein feature.
5：Data Analysis Methods:
The equal error rate (EER) is adopted to be the evaluation criterion.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
