Lightweight CNN and Image Enhancement Using in Palm Vein Recognition

Information safety is becoming increasingly important in modern society, and biometric recognition technologies have become a key to achieve information security due to its universality, permanence, distinctiveness, and collectability. This research proposed a palm vein verification system based on CNN (convolution neural network). In preprocessing step, we first perform ROI localization and image contrast adjustment, and then proceed to feature extraction through lightweight convolutional neural network. Finally, we validated our method on the CASIA palm vein dataset. As a result, the system achieved an Equal Error Rate (EER) of 0.529%, which is superior compared to other renowned architectures, and it also possesses the minimal parameter count and the fastest execution speed.