A transfer Learning features for fingers recognition

Abstract

Person identification security is one of the most pressing challenges in current times. There is a high need for a trustworthy and secure identity verification solution. A biometric identification system can be a safe and secure way to identify someone. Because it has the ability to identify individuals,in other way we find the finger knuckle print (FKP) and finger vein (FV) is considered two of the developing hand biometrics. In our experience, we used convolutional neural networks as one of the basic structures for deep learning because they are the best in image analysis. With Using transfer learning ( Pre-trained CNN models) we will test a biometric system based on fingers knuckle and finger vein.with two datasets of finger-knuckle-print and finger-vein, containing 11736 images of finger knuckle and finger vein (7920 knuckle, 3816 vein). In addition, we will apply a multimodal system, by using the fusion of the unimodal scores in order to improve the performance and get better results. The experimental results were very encouraging and showed the potential for biometric applications utilizing the finger knuckle print and finger vein. The thesis also performs a comparison of the identification performance of the system with different CNN models to choose the best result for the knuckle and vein recognition task,we detail all the results of our tests in this thesis.