Font Size: a A A

Key Technologies Research On Finger Vein Recognition For Low Quality Images

Posted on:2015-01-11Degree:DoctorType:Dissertation
Country:ChinaCandidate:T LiuFull Text:PDF
GTID:1108330509461071Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Finger-vein recognition can be used for personal identification according to the pattern differences among different finger-veins. It is widely used in security field, and becomes a hot topic in the field of biometric technology, because of the good performances on anti-spoofing and safety. The main challenge facing in the field of finger-vein recognition, currently, is how to improve recognition performance for low-quality images. Key techniques of finger-vein recognition, including finger-vein representation, image segmentation, finger-vein pattern restoration, and feature matching, are deeply studied in this dissertation. Some research achievements can improve recognition performance for low-quality images, and can also be used for relevant fields, such as fingerprint recognition, palm-vein recognition, etc. The main contributions and innovations are summarized below.(1) Points out that line-structure of finger-vein is the base of finger-vein recognition because of its distinctness and stableness for distinguishing different persons and recognizing the same person, after analyzing the finger-vein structures from low-quality images; proposes four principles, including caliber uniformity, node replication, loop splitting, and virtual connection, to extract line-structure of finger-vein; proposes a BSBT(B-Spline Binary Tree) model, which uses binary tree to describe the relationship of different finger-vein branches, and uses B-spline to describe the spatial structure of each finger-vein branch. This model can fully describe the distinct and stable structure of finger-vein, and help to improve recognition performance.(2) In the aspect of image segmentation of finger-vein, an ARLT(Adaptive Repeated Line Tracking) method is proposed, for overcoming the drawbacks on adaptive ability and efficiency of the classical RLT(Repeat Line Tracking) method. The new method improves the adaptive ability for finger-vein images with differences on vein caliber and image contrast, and decreases time consumption caused by invalid line tracking. This method can extract precise details of depicted veins from low-quality images, and inhibit the reduction of real finger-vein’s distinctness, for settling the bases of reliable recognition of finger-vein.(3) Fracture and burr always exist in the finger-vein patterns extracted from low-quality images, which lead to a decline of recognition performance. A pattern restoration method based on BSBT model is proposed. The new method locates the positions of fracture and burr based on the leaf nodes of BSBT model; restores fractured branches according to three constraints of direction, variance and boundary, and deletes burr according to the node and length attributes of BSBT model. This method can improve the stableness of line-structure of finger-vein, and help to improve recognition performance.(4) For solving the problem of low recognition performance caused by damaged vein patterns, a feature matching method based on BSBT model is proposed. According to the characteristics that BSBT model can independently describe each vein branch, the new method executes segment matching and comprehensive judgment operations to reduce the influences on feature matching from damaged vein patterns, and in the end improves recognition performance for low quality images.
Keywords/Search Tags:biometric features, personal identification, finger-vein recognition, image representation, image segmentation, image restoration, pattern recognition
PDF Full Text Request
Related items