On the Performance of Pretrained CNN Aimed at Palm Vein Recognition Application

Abstract

Biometric technology has been very highly developed as a recognition system as personal identity. Because biometric is attached to human body such as physical or behavioral. Many applications adopt biometric recognition as their security and access system such as smart house or smart building, banking access system, cellular phones and many more. Vascular pattern include vein pattern is being a very fast-growing research. Vein pattern identifies an individual from his vein features. The quality of infrared vein images need to be enhanced by increasing the contrast to extract the object from the background Many methodologies has been developed to create a robust system of recognition from feature extraction to classification method. And high developed algorithm for classification which is rapidly being developed is deep learning, Convolutional Neural Network (CNN). There are four pretrained structure of CNN that discussed in this paper, AlexNet, VGG-16, VGG-19 and GoogLeNet. AlexNet seems to be the simplest in depth. The accuracy of AlexNet is better among others with 93.92% ±0.98334.