Finger Vein Recognition with Channel Spatial-enhanced Multiscale Feature Fusion Deep Convolutional Neural Network

FInger-vein Finger-vein recognition belongs to the field of biometrics, similar to other types such as faces, fingerprints, irises, and palmprints. Finger vein recognition is becoming an essential means of identification because of its uniqueness, live detection, security, and many other advantages. The recognition performance fails with degradation factors including blur, noise, and nonuniform illumination that exists in the image while acquiring the finger-vein image requires image restoration. However, previous Models produced lower image quality and slow inference speed. To overcome the challenges in the existing approaches, Channel Spatial enhanced Multiscale Feature Fusion Deep Convolutional Neural Network (CSMF2-DCN) is proposed in the research. CSMF2 is the proposed attention mechanism implemented in the research model to avoid overfitting and improve recognition efficiency. The features are extracted using pre-trained deep local weighted features, which aggregate several approaches that aid in retaining the performance of DCN to recognize finger veins. Finally, the CSMF2-DCN model attains high accuracy in finger vein recognition compared to other conventional research. The performance of the proposed method attains $\mathbf{9 6. 4 4 \%}$ accuracy, $\mathbf{9 6. 4 1 \%}$ precision, and $\mathbf{9 6. 4 6 \%}$ recall, which shows the high efficiency of the research model.