Study On Finger Vein Recognition Algorithm

As people pay more and more attention to information security,the secure identification method is more and more favored.Finger vein recognition technology has become a hot research topic among scholars at home and abroad because of its advantages of high anti-counterfeiting,uniqueness,stability and live detection.At present,most of the algorithms based on vein structure feature only consider point features such as vein intersections and endpoints,but ignore the curve features of veins,which will result in the loss of part of the structural information and affect the recognition results.And because there are few intersections and endpoints of vein images,some low-quality finger vein images and the limitations of segmentation algorithms will reduce the effectiveness of vein point features,resulting in unsatisfactory recognition results.So it is not enough to consider only point features such as vein intersections and endpoints.Aiming at these problems,this paper mainly conducts the following researches:(1)A finger vein recognition algorithm based on curve descriptors is proposed.First,the skeleton structure of the finger vein is extracted,the intersections and endpoints of the veins are detected,and the vein skeleton is divided into several curved segments using the intersections and endpoints.Secondly,the curve arc descriptor and the intersecting arc descriptor are proposed based on the relative position and shape characteristics of the intersection and the curve segment,and the structure feature matrix of the finger vein image is extracted according to the curve descriptor.Finally,according to the proposed weighted distance formula,the matching intersecting arc pairs are calculated for feature matching and image matching.(2)The four algorithms of LBP,LTP,SURF and finger vein recognition based on curve descriptors were used to conduct experiments on the self-made finger vein database collected in the laboratory,and the experimental results were analyzed.The experimental results show that the equal error rate of traditional LBP,LTP and SURF algorithms is4.46%,3.98% and 6.06% respectively,while the equal error rate of the finger vein recognition algorithm based on curve descriptors proposed in this paper is only 1.63%,far lower than that of LBP,LTP and SURF algorithms.It shows that the method presented in this paper has a certain universality and application prospects in finger vein recognition.