Research On Technology Of Contact-less Palm Vein Recognition System

With the popularization of Internet applications and human dependence on the Internet,the security of personal information is greatly threatened.How to accurately identify personal identity has been imminent.As a new biometric identification technology,palm vein identification has become a high-security identification method due to its uniqueness,living body identification,and difficulty in copying.At present,most palm vein systems require users to place their hands on the image acquisition plane when acquiring data,and at the same time use a positioning device to limit the posture of the palm.This acquisition method will cause problems in hygiene and health.Therefore,non-contact palm vein identification without positioning device came into being.However,the non-contact acquisition method leads to a larger change in the posture of the palm,which poses a huge challenge to the subsequent preprocessing and the location of the reference point in the region of interest,and ultimately leads to a rapid decline in the recognition accuracy.This thesis focuses on the non-contact palm vein image acquisition,preprocessing,ROI reference point positioning and identification algorithm problems.The main work of this thesis is as follows:1.Based on the hardware module OVS-200,the non-contact palm vein video image acquisition system is design and implemented,and the data set acquisition is completed.The system can perform soft positioning of the palm position,which to a certain extent makes up for the shortcomings of the non-contact palm vein hardware module without a positioning device.First of all,the system fixes the palm in the center of the field of view through the palm positioning algorithm and before acquisition,which can limit the position and posture of the palm to a certain extent and reduce the difficulty of identification.Secondly,the system records the information of each palm in detail through the database,which provides data support for subsequent algorithm research and system implementation.Finally,the system is equipped with perfect collection guidance and prompts,including user mode and administrator mode.Users can complete the collection work by themselves,which greatly liberates manpower.The hardware collection end of the system is small and portable.The administrator can also view and manage user information through the database later.And this system is highly user-friendly.2.According to the characteristics of the non-contact palm vein image acquisition device,a new method for locating the region of interest(ROI)reference point is proposed,which corrects the image offset brought by the acquisition process and improves the positioning accuracy of the palm ROI.The key reference points are detected and positioned on the palm image through the HOG feature and the SVM classifier,the angle of the palm image could be corrected by the reference point,and the coordinate system can be established to obtain the ROI image.The ROI extraction rate of the proposed method reaches 96.78%.Secondly,for the system,a palm detection module is added in front of the ROI reference point positioning module to accelerate the non-contact palmar vein recognition rate by screening out unqualified palm video frames.Finally,by improving the existing SIFT feature method,the concept of feature template is proposed and a lower misrecognition rate is achieved.The advantages and disadvantages of two feature extraction methods,NBP and SIFT,were compared through experiments.And the feature point screening of the SIFT method with good performance was improved based on the template matching method.The concept of constructing feature templates was proposed,and then the number of templates was tested,the experiment has proved that the EER of the system is stable at 0.43%when the number of templates exceeds 4.3.Aiming at the above algorithm research of palm vein image recognition based on OVS-200 equipment,the corresponding contact-less palm vein recognition system software has been designed and implemented.And the real-time performance of the recognition system has been tested as well.First of all,the system can realize the registration and identification of multiple people,and use the SQLite database to store the palm feature template data extracted when users register.Then the running rate of the system is tested,the total time for the system to complete a recognition is 151.46ms.Finally,the time test was carried out for each module of the system separately,and the distance test is carried out for the reference point positioning part that takes the longest time in each part of this system.These experimental results show that the closer the distance between palm and device,the faster the reference point positioning speed.The recognition rate of the system is also faster.