Image Mosaic And Recognition Techniques For Finger Veins

With the development of mobile Internet and intelligent technology,the requirements for information security technology are getting higher gradually;also,more reliable biometric technologies are expected to be used for identity authentication.At present,the fingerprint recognition technology is relatively mature,which is widely used in a variety of small and medium-sized intelligent devices.However,fingerprint recognition technology has a fatal weakness,that is,fingerprints can be mocked easily.As a new and high-security biometric technology,finger vein recognition realizes identity authentication by use of finger vein characteristics.Since finger veins are in vivo and they are more difficult to be mocked,they have high degree of security.Theoretically,it is possible to realize the miniaturization of the finger vein recognition system and to be designed in small-sized intelligent devices in the future.Therefore,this thesis is mainly focused on the finger vein image acquisition,finger vein image mosaic,finger vein feature extraction and matching.The main works are as follows:In order to solve the problem that miniature finger vein image acquisition module cannot obtain adequate finger vein information,image mosaic technique was employed in this thesis.It can achieve finger vein image acquisition device miniaturization using the mosaic technique to ensure adequate finger vein information to improve the recognition accuracy and reliability.Controllable finger vein image mosaic algorithm based on Retinex was proposed to achieve two homologous finger vein image mosaic.Finger vein image mosaic technology provides technical support for miniaturization of vein image acquisition devices.Aiming at getting the finger vein minutiae points which could be chosen as feature points for subsequent matching,finger vein image segmentation and vein line refinement were investigated.After using the finger vein pattern extraction method based on directional filtering,the correspondence between Hessian matrix eigenvalues and image linear structures was used,and the algorithm of finger vein image segmentation based on Hessian matrix was proposed.By comparing finger vein image segmentation results using the two methods,it can be found that the latter method is able to extract finger vein lines with different width.In the present work,finger vein image acquisition module was developed,which could be used for finger vein image mosaic and finger vein recognition;The GUI program was designed to control the progress of finger vein image acquisition,image mosaic,image recognition and image processing by using Matlab commercial software.In order to achieve the finger vein feature extraction and recognition,the matching between present finger vein image and the spliced finger vein image was studied.In this thesis,the method of feature point matching based on Hausdorff distance was used to match two finger vein images.The matching method based on SIFT was used as well.Subsequently,these two algorithms were verified experimentally,and the homologous matching and heterogeneous matching between self-built finger vein image feature library and finger vein image verification database were carried out.After comparing the results of these two methods,it could be found that the recognition accuracy of the latter method is higher;also,its adaptability for the vein image translation and rotation is better.