Multispectral Palm Vein Recognition Using Hierarchical Approach and Feature Fusion

Now-a-days biometrics plays an important role in authentication. Biometrics makes authentication process automatic. The advantage is that the user need not carry or remember anything. Palm vein is superior to other biometrics because it cannot be forged. In earlier research, usually the images captured in natural light were used. It cannot provide an acceptable recognition performance for contactless palm vein images. In order to obtain the high recognition rate with more particular information, here multispectral Palmprint images are used instead of natural light palm images. Firstly, region of interest (ROI) of palm is taken and image enhancement is done to enrich the vein texture, secondly the features are extracted based on Block Dominant Code (BDOC) as a rough feature and Block-based Histogram of Oriented Gradient (BHOG) as a fine feature. Thirdly Hierarchial classifier is used to find the matching between two palm features. Finally, different feature from different bands are fused to improve the recognition rate. Experimental result proves that the accuracy rate of the CASIA multispectral Palmprint database is superior to the previous performance of CASIA Palmprint database.