Research On Finger Vein Recognition Algorithm

As our society steps into an era of internet of things nowadays,information security is becoming more concerned.The disadvantage of traditional identity authentication methods based on passwords or storage medium is emerging gradually due to the contradiction between security and usability.The emerging biometric identity authentication technology solves this problem and greatly improves the security and convenience level of the authentication process.In recent years,finger vein recognition technology has become a research hotspot in the field of biometric-based identification technologies with its excellent performance.It is widely used in the field of information security because of its difficulty in copying and comfortable interaction.The dissertation carries out research on finger vein recognition algorithm,analyzes the problems in its key links and puts forward the corresponding improvement algorithms.Our work in this thesis can be summarized into the following four aspects.(1)This thesis introduces the background of biometric identification technologies,summarizes several typical biometric identification technologies and compares their advantages and disadvantages.Then,it outlines the concept and advantage of finger vein recognition technology and introduces the present development at home and abroad.(2)This thesis proposes a finger vein acquisition device model based on an adjustable light source.For the problem of nonuniformly illuminated images captured by the near-infrared light source with single intensity in the traditional finger vein acquisition device,an improved finger vein capture device model and an adjustable light source algorithm based on area division are designed.The improved device model based on adjustable light source designed in this study can be applied to objects with different vein characteristics.The quality of finger vein images acquired by this acquisition is improved.(3)Two finger vein image preprocessing methods are put out for the finger vein images in SDUMLA-FV database.Aiming at the problem of the inapplicability of traditional edge detection algorithms to finger vein images,an adaptive edge detection algorithm based on potential pixel points is proposed.Aiming at the problem that there may be a discrepancy between different finger vein images from the same finger with the traditional ROI(Region of Interest)extraction methods,a novel ROI extraction method based on the distal inter-joint line for finger vein images is proposed.The designed preprocessing methods can accurately locate the edge of the finger and extract the region of interest,which are helpful to improve the accuracy of the subsequent recognition algorithm.(4)This thesis proposes a new finger vein recognition algorithm.By analyzing the limitations of local binary model and its improved models in feature extraction,a new local binary model,called local binary pattern based on local macrostructure and microfeature fusion(LMMF-LBP),is proposed to characterize vein features more accurately.Then,Finger vein recognition algorithm based on LMMF-LBP(LMMF-WPLBP)is designed using partition algorithm and weighted template.The results of experiment show that the finger vein recognition algorithm based on LMMF-WPLBP can not only improve the recognition performance,but also has strong robustness.