Research On Near-infrared Contactless Palm Datum Point Locating And Palm Vein Recognition Algorithm

With the advancement of science and spread of new technologies such as cashless payment.how to reliably,rapidly and accurately identify different people has become an urgent problem to be solved.Due to the impact of the epidemic of covid-19,in order to prevent viruses,bacteria and other microorganisms from forming cross-infection between people through biometric collection equipment,traditional contact biometric identification technology has gradually been unable to meet the needs of current generation.Secondly,since the traditional face recognition,fingerprint recognition has the risk of being easily stolen and impersonated,which can no longer meet the requirements of high security.Palm vein is a new biometric identification technology emerging in recent years,which has the advantages of being difficult to tamper with and easily accepted by users.In this paper,in the contactless palm vein dataset collected without background control,the palm datum point locating,region of interest extraction algorithm and palm vein feature matching algorithm are studied.By comparing the method based on segmentation often used in former studies,the datum point extraction algorithm based on Support Vector Machine(SVM)feature classification proposed in this paper,in the case of no background control,can more effectively and accurately determine the finger seam datum point.Then,a scale-invariant feature transform method based on similarity statistics was proposed to eliminate mismatches,which optimized the traditional SIFT matching method according to the characteristics of the region of interest of palm vein acquisition.Then,the method of neighbor-based binary pattern and proposed mismatch elimination method was used to compare the accuracy of the matching.Finally,a fusion algorithm that combines NBP features and SIFT-based feature mismatch elimination is proposed and the accuracy of palm vein matching was improved when the two methods were combined.The main work and research results of this dissertation are as follows:1.Collected and established a contactless palm vein database CNPVD including 400 people without background control and collected at different distances,the algorithm is tested through this dataset which includes more than 6000 high-definition palm vein images.2.The experiment verifies the shortcomings of the threshold-based segmentation method in extracting datum point and the region of interest from the contactless palm data without background control.Then,an algorithm based on SVM for localization and extraction of palm region of interest is proposed.Experiments on the CNPVD dataset show that the method has a higher success rate in obtaining datum points and improved performance compared with the method based on threshold segmentation.3.According to the characteristics of palm vein image matching,a SIFT mismatch elimination algorithm based on similarity statistics is proposed,and its performance is tested.The performance of the method is tested on the palm region of interest collected in the CNPVD dataset.Experiments show that this method can effectively reduce the occurrence of mismatches and improve the accuracy of matching.4.A fusion algorithm that combines NBP features and SIFT-based feature mismatch elimination is proposed,and the effect of combining the two methods on palm vein alignment is tested.The experimental results show that when the two methods are fused together,the FAR and FRR can be guaranteed within an acceptable range,and results are also achieved in the CNPVD dataset test.