Research On Palm Vein Recognition Algorithms

With social and economic developments, especially the popularization of Internet applic-ation, the requirements for information security have been sharply increasing. How to identif-y a person’ identity accurately and reliably, has become a problem to be solved. However, tr-aditional authentication methods, such as passwords and keys, are unable to satisfy these ex-pectations, because they are prone to be forgotten, lost or stolen, can not meet the requireme-nt for high security. Biometric recognition is a effective way to solve the problem. Palm vein recognition which rises in recent years has high security, high accuracy and easy acceptanc-e, etc.. This new biometric recognition has been paid attention by many academics and ind-ustries, and has a broad development potential and market prospect. In this thesis, the palm v-ein recognition process is further studied, mainly in the ROI extraction, image enhancemen-t, features extraction and matching method and multi-biometric based on palm vein, and s-ome new recognition algorithms are proposed. The main work and research contributions o-f this thesis are as follows:1. The palm vein ROI extraction method is studied. At present, the pam vein ROI extr-action method is to mostly extract a rectangular area from the palm center. An improved m-ethod to extract a circular area as ROI is proposed. The experimental results show that thi-s method is effective to extract the palm vein ROI, and the distinction is high between imag-es.2. Palm vein image enhancement method is studied. An sharpened enhancement metho-d based on spatial domains is proposed in this thesis, the experimental results show that th-e method significantly highlights the coner features of palm veins.3. Palm vein recognition based on local invariant feature extraction is studied, and two kinds of palm vein recognition algorithm based on local invariant feature extraction is proposed: a recognition algorithm based on ORB feature extraction, and a recognition algorithm based on AGAST feature extraction and BRIEF description, and get a good recognition effect. The method of eliminating false matched feature points is studied, and a method based on matched feature points offset is proposed to eliminate mismatches, and the number of false feature points matches is reduced.4. Palm vein recognition based on image texture analysis is studied, and two pal-m vein recognition algorithms based on Gray Level Co-occurrence Matrix and Local Binary Pattern are proposed5. The palm vein recognition based on mlti-algorithm fusion is studied, and two recognition algorithms are proposed. One is based on SIFT features and ORB feature extraction and score-level fusion, and two dynamic weighted score-level fusion conbin-ation rules are proposed to enrich the conbination rules. Another is based on AGAST and LBP and support vector machine. Palm vein recognition rate is improved effecti-vely.6. A feature-level fusion method based on feature vector similarity is proposed. The p-alm vein recognition based on multiple-sampling and feature-level fusion obviously impro-ves the recognition rate, even under the situation that the hand pose largely changes, the reco-gnition algorithm obtains a high recognition rate.As a new biometric recognition technology, palm vein recognition just rises in recent y-ears, and has a great development potential and market prospect. It is worth doing more in-depth research about palm vein recognition.