Study On Infrared Image Recognition Method Of Palm Vein

In modern society,identity recognition technology is undergoing rapid update and development to adapt to the ever-increasing demand for information security.How to ensure that personal funds and information are not stolen and leaked has become a problem that all levels of society pay close attention to.In view of the shortcomings of previous identification methods,such as easy loss and easy forgery,biometric authentication methods have been developing vigorously in the new era of identity authentication because they take advantage of their own physiological characteristics.Among them,the identification technology based on the palm vein has been gradually applied in the society because it is difficult to forge.This thesis is mainly about the infrared image recognition method of palm vein.Aiming at the low quality of the collected palm vein images,this thesis first preprocesses the palm vein images,and then extracts and matches the features of the images.In the pre-processing stage,the image is firstly processed by NSCT(non-subsampled Contourlet Transform)to obtain low-pass sub-band images and band-pass sub-band images in different scales and directions.Linear enhancement of low pass sub-band coefficients can improve the image contrast.For the band-pass sub-band coefficients,based on the energy distribution,the noise threshold of each sub-band is determined adaptively,and a weak edge enhancement algorithm is proposed to enhance the detail texture and suppress the noise.After the above steps are processed,the enhanced palm vein images are segmented and refined.The image segmentation method used in this thesis is based on regional active contour model,which can carry on the image segmentation along the vein texture details and retain the local information.Moreover,this method can segment the vein images with uneven intensity distribution and noise.In the stage of feature extraction and matching,this thesis studies the extraction method of feature points based on structural features,including the method based on venous endpoint and bifurcation point and the method based on SIFT(Scale Invariant Feature Transform)algorithm.The method based on venous endpoint and bifurcation has the advantage of short recognition time,but the correct recognition rate is low.The method based on SIFT algorithm has strong robustness in image rotation,scaling and view angle transformation,but the recognition time is long,and the real-time performance is poor.According to the advantages and theirdisadvantages of the two methods,this thesis combines the above two methods and proposes a palm vein recognition method based on quadratic matching.The above palm vein recognition method has been carried on the experiment,and the results show that the combination of the two methods can effectively reduce the recognition time under the premise of ensuring the recognition accuracy.From the experiment,the thesis gets four evaluation indexes,the correct recognition rate,false acception rate,false rejection rate and recognition time,which are 97.15%,1.62%,1.23%,2.17 s respectively.The experimental data shows that the recognition method has practical application value.