In many real applications,compared to unimodal biometrics,multimodal biometrics always behave better in security and universality.Multimodal biometric-based technology therefore has been an attractive topic in biometric research community.In addition,multimodal finger biometrics has been a significant research direction in biometrics because of the high portability of finger itself.The quality of finger-vein(FV)images collected by existing multimodal finger imaging devices is poor,and the regions of interest(ROIs)localization of finger-vein images and finger-knuckle-print(FKP)images is unreliable,which is not beneficial for multi-modal feature fusion of fingers.In order to improve the quality of finger-vein imaging and the accuracy of locating the regions of interest(ROIs)in FV images and FKP images,a new finger imaging system and a new ROI localization method are both proposed.And one encoding fusion method based on local Gabor binary pattern(LGBP)features is selected to achieve multimodal finger-feature fusion.The main contributions are as follows:(1)A finger arch imaging model with multi-spectrum and multi-intensity lights is designed.This proposed model can reliably improve the quality of finger-vein images.(2)A new localization method for the ROI of FV and FKP is proposed.The inflection points of finger arches are used for finger-knuckle-print ROI localization.Based on the spatial relationship between two imaging spaces of a binocular camera,the coordinates of FKP ROIs are mapped into the FV imaging space for delineating FV ROIs.Experimental results show that the proposed methods are more desirable in dual-mode finger imaging and FV-FKP ROI extraction.And the designed identity recognition system can realize multimodal finger-feature fusion effectively and achieve higher recognition accuracy and efficiency. |