PROJECT TITLE :

Deep Representation based feature extraction and recovering for Finger-vein verification - 2017

ABSTRACT:

Finger-vein biometrics has been extensively investigated for personal verification. Despite recent advances in finger-vein verification, current solutions fully depend on domain data and still lack the robustness to extract finger-vein options from raw pictures. This paper proposes a deep learning model to extract and recover vein options using limited a priori knowledge. 1st, primarily based on a mix of the known state-of-the-art handcrafted finger-vein image segmentation techniques, we have a tendency to automatically establish two regions: a transparent region with high separability between finger-vein patterns and background, and an ambiguous region with low separability between them. The first is associated with pixels on which all the above-mentioned segmentation techniques assign the identical segmentation label (either foreground or background), whereas the second corresponds to all the remaining pixels. This theme is used to automatically discard the ambiguous region and to label the pixels of the clear region as foreground or background. A coaching information set is constructed primarily based on the patches centered on the labeled pixels. Second, a convolutional neural network (CNN) is trained on the ensuing knowledge set to predict the chance of each pixel of being foreground (i.e., vein pixel), given a patch focused on it. The CNN learns what a finger-vein pattern is by learning the difference between vein patterns and background ones. The pixels in any region of a take a look at image will then be classified effectively. Third, we tend to propose another new and original contribution by developing and investigating a absolutely convolutional network to recover missing finger-vein patterns within the segmented image. The experimental results on 2 public finger-vein databases show a vital improvement in terms of finger-vein verification accuracy.