| The research on identification and verification of palmprint and palmvein started relatively late, whereas it is progressing remarkably fast due to multi-fold advantages.Palmprint systems are of low cost, good user acceptance and high accuracy. Since hypodermal palmvein features are not easy to leak or forge, the privacy and anti-spoof capacity are improved.Unfortunately, with the advancement of theoretical research and the continuous promotion of application fields, the open issues in conventional biometrics(including palmprint and palmvein) systems are exposed, which seriously restrain the development and popularization of related technologies. The open issues are as follows.(1) In contact-type acquisition system with low user acceptance, the device surface will be polluted to degrade the acquisition quality. The contact will also lead to hygiene problem and risk of infectious diseases.(2) Single-modal biometric cannot satisfactorily meet all the requirements,including high and stable verification accuracy, widened availability and application range, high degree of freedom, and strong robustness against spoof attacks.(3) Biometric features are immutable and scarce, which implies that biometric templates cannot be revoked, updated or reissued, even if they are compromised.In order to solve the aforementioned problems and exploit the advantages of biometric fusion, this dissertation focuses on palmprint and palmvein, and also introduces cancelable biometrics technology. In contactless acquisition environment,fusion verification of palm-print-vein is studied. The main contributions and innovation points are as follows:(1) Design and implementation of contactless palm-print-vein acquisition systemContactless acquisition is more flexible and of high acceptance; however, the acquisition condition is more complicated. Through the discussion on the properties of dual-source palm-print-vein acquisition, a contactless palm-print-vein acquisition system with practical framework was designed and implemented, whereby a palm-print-vein database was established, which was used for the evaluation of the proposed fusion verification algorithms.(2) Region of interest(ROI) location based on point set operation(PSO)In contactless palm-print-vein acquisition, several technical challenges are incurred by several unstable factors. ROI location based on PSO fully utilizes the direction of palm. The segmented palm region was translated along three perpendicular directions,and then the intersection between the three translated images and original image was computed. ROI was accurately located with post-processing. PSO-based method can better resist the interference of background and palm position. Moreover, compared with competitive hand valley detection(CHVD) method, PSO-based method reduces the searching area, so it saves computational cost and improves location accuracy.(3) Cancelable palm-print-vein fusion verificationFusion of palmprint and palmvein was implemented in cancelable domain,whereby cancelable transform was embedded in the fusion framework. In order to avoid the complexity increment of computation and storage, the cancelable dual-source templates were selected according to some rules before fusion, which also reduced the leakage of original features. Two template selection rules, dubbed “verification accuracy sequencing” and “small covariance selection”, were proposed and compared.Compared with the single-modal verification of palmprint and palmvein, the verification accuracy is improved by the novel technology, “cancelable palm-print-vein fusion verification”; while the privacy does not deteriorate.In summary, this dissertation studies the preprocessing of contactless palmprint and palmvein as well as the corresponding cancelable template generation schemes. The fusion verification framework of contactless cancelable palm-print-vein improves the privacy, security and verification accuracy, which is of remarkable theoretical significance and important engineering application values. |
PDF Full Text Request