Dorsal Hand Vein-Biometric Recognition Using Convolution Neural Network

In this paper, a convolution neural network (CNN)-based vein recognition approach is used for dorsal hand vein patterns. Apart from using pre-trained version, VGG Net-16 model is fine-tuned on four datasets of dorsal hand vein images (good quality, medium quality, and low quality) and augmented images (between the two images from genuine matching or false matching). All four datasets consist of dorsal hand vein images of left and right hands. The comparison of results proposed model is done with other CNN models like VGG Face and VGG-19 (with and without fine-tuned) along with a recent work based on transfer learning. The accuracy of recognition of proposed work using fine-tuned VGG Net-16 model obtained is 99.60% for good quality images; for medium quality images, it is 98.46%, and 97.99%; for low-quality images, it is found to be.