Finger Vein Recognition Algorithm Research

With the rapid development of information technology,information security has become increasingly important.Finger vein recognition as a new type of biometric technology has been favored by more and more experts and scho lars.Finger vein recognition has the following advantages:(1)High security: Traditional biometric technologies,such as fingerprint recognition and face recognition,use human body surface information for identification.The finger vein recognition uses the information of the finger vein in the body to identify it,so it is more difficult to be stolen and copied.(2)Non-contact collection: It can avoid indirect contact with different people,and it is more hygienic and convenient.(3)Live acquisition: The finger vein image is based on the blood flowing in the human finger can absorb the light of a specific wavelength(720 ~ 1100 nm near infrared),the image formed by the reflection of the finger.Only the living blood can absorb specific wavelengths,so the finger veins are a kind of live collection and thus more secure.Finger vein recognition is divided into six processes: image quality assessment,image rotation correction,image region of interest extraction,image enhancement,feature extraction,and recognition matching.About finger vein image quality assessment.Some of the lower-quality images appear during finger vein acquisition,and these images have reduced recognition performance due to less effective features.This paper proposes the differe nce in the depth of the venous line between the vein image of good and poor finger veins and the complexity of the vein line structure of the vein vein of the finger vein.It is the most critical clue of the image quality assessment.It can be concluded that the number of vein points can be used to compare the images.to evaluate.First extract the gray value of the vein area.Then the number of vein points was detected by the depth threshold,and the normalized mass fraction was finally obtained by the number of vein points,thereby realizing the evaluation of the image quality of the finger veins.Experiments show that the database with finger rotation,the EER without image quality analysis is 3.31%,the image quality analysis(AND rule)EER is 2.15%,and the database without finger rotation,no image quality analysis EER is 1.02%.The image quality analysis(AND rule)EER was 0.51%.About finger vein image rotation correction.Existing finger vein image preprocessing methods often ignore the effect of finger rotation on recognition,or the effect of processing is not ideal.In order to overcome this problem,this paper proposes a method based on the minimum circumscribed rectangle for image rotation correction.First,the finger edge image is obtained based on the dual threshold detection method.Then,based on a fast algorithm for extracting the minimum bounding rectangle of the target image,the minimum circumscribed rectangle of the finger is found.The image is then rotated based on bilinear interpolatio n.Then based on the sliding window to find out the position of the finger joint,combined with the position of the finger joint and the size of the minimum circumscribed rectangle,the position and size of the region of interest are determined.Experiments show that the EER is 1.40% with the rotation-corrected MHD method.There is no choice of correction MHD method,EER is 3.31%About the enhancement of the finger vein image.Although the image quality of the finger veins has been evaluated and screened,the contrast of the images can be further enhanced by the image enhancement technology,thereby improving the recognition accuracy.This paper presents a method of finger vein image enhancement based on fuzzy fusion.Firstly,two images are generated using Gabor filtering and Retinex filtering in four directions.Then,different fusion rules are compared experimentally.The performance of fuzzy FOM based on the minimum rule is the best,and the database EER without finger rotation is 0.81%.About finger vein feature extraction.Finger vein feature extraction is a key step in the extraction of finger vein features.The recognition rate of single feature extraction is low.This feature-level fusion method extracts features,in which maximum curvature,repetitive line tracking,wide line detector,main curvature,Gabor filter are used.The features of the isotropic non-abstract wavelet transform feature extraction and the majority vote,weighted average,STAPLE,STAPLER,and COLLATE fusion techniques were used to fuse the feature levels.Experiments were conducted to compare different fusion methods.The best performance was achieved when all 6 functions(MC,PC,WLD,GF,RLT,and IUWT)were fused using MVs,and the database EER without finger rotation is 0.26%.About finger vein image recognition.The finger vein image recognition is the last and the most critical step.This paper uses the MHD-based algorithm to recognize the finger vein details.The intersection points,endpoints and the two combined methods were combined with the previous method.The EER was 1.21% and compared with other methods,it was superior to other methods.