Research On Extraction Of Region Of Interest For Finger Vein Recognition System

With the rapid development of information science and technology,how to easily and reliably identify and protect the security of one's own information has become an important issue that needs to be urgently addressed in today's society.Finger vein recognition technology has attracted more and more interest from biometric researchers due to its unique advantages such as high security level,biometrics,and internal characteristics.Finger vein recognition system mainly includes four stages of image acquisition,preprocessing,feature extraction and recognition matching.During the preprocessing step,it is important to accurately extract the region of interest(ROI)of the finger vein image,which directly affects feature extraction.And identify matching effects.To extract the ROI of the finger vein image in the finger vein recognition system,the classic method is to obtain the contour of the finger through edge detection algorithms such as Sobel and Canny,and then cut out the area of interest based on the contour.The finger vein image contains a lot of device background noise and other random noise,so the effect of edge detection is not ideal.In order to be able to extract the region of interest of the finger vein image robustly and accurately,the following work is done in this paper:(1)This paper proposes a method for extracting regions of interest of finger vein images based on template edge detection.It uses template edge detection to obtain rough edges of fingers,and uses fully-connected neural networks to repair wrong boundaries.This method can accurately extract the finger edge information of low-quality finger vein images,and then cut out the region of interest.Through comparative simulation experiments,it is verified that the algorithm can accurately extract the region of interest of the finger vein image,and greatly improves the accuracy of the finger vein recognition system.(2)There are several pictures in the finger vein data set.Due to the overexposure,the edge of the finger is fused with the background of the device.It is difficult to accurately obtain the edge information based on the method of edge detection.Therefore,this paper proposes a method for extracting regions of interest in finger vein images based on deep learning,using the classic segmentation framework Mask-RCNN in the field of deep learning to detect targets in the finger vein images,that is,fingers.In this paper,according to the characteristics of the finger vein data set,the anchor box definition of the RPN network in the framework was redesigned,and an auxiliary loss function was added to the mask prediction branch to help complete the mask prediction.The experimental results show that this method can accurately extract the region of interest of the finger vein image with extremely poor quality,which further improves the accuracy of the finger vein recognition system.