Deep learning-based Region of Interest Extraction for Finger Vein Images

Abstract

Biometric technology has played an important role in our lives. Finger vein recognition has many advantages as a new biometric technology. Different from iris, fingerprint or face, finger vein recognition is cheap, Non-contact, living. However, finger vein recognition is susceptible to many factors such as light environment, finger moving, etc. In order to solve these problems, in this paper we first we first propose a method for extracting region of interest for finger vein images based on a convolutional neural network. Secondly, we have integrated several common finger vein databases and manually labeled the region of interest. We also extended the dataset using data augmentation methods. Finally, for the transformed images, we calculate the IOU of the prediction box and the theoretical box to measure the robustness of the method and Experiments show that our method improves the performance of the recognition system.