Design And Realization Of Multi-camera Finger Vein Recognition System

With the popularity of smart terminals,people put more attention to the security of information.There is a trend that the new generation of biometric identification will replace the traditional authentication.As a representative of biometrics,finger vein recognition is safer and more accurate than other biometric identification.However,most products use a single camera for image acquisition,which directly results in the large distance of the finger vein imaging,thus increasing the height of the finger vein recognition device.Therefore,it is urgent to design a miniaturized finger vein recognition system to meet the needs of most requirements.In this thesis,a large area of finger vein images are collected in a limited space by multiple CMOS sensors and the venous features are extracted and matched on the embedded software platform.The design not only realizes the miniaturization of finger vein recognition system,but also has high real-time performance and accuracy.The work in this thesis is as follows.(1)In order to solve the problem of limited view angle of a single camera,a finger vein imaging device with multiple CMOS sensors is designed.On this basis,the camera driver was developed to capture the whole finger vein.(2)To improve the efficiency of the whole system,the algorithm of finger vein recognition is optimized.The auxiliary table is used to reduce the noise the image quickly.The pixels of ranks and columns are combined to accelerate the registration of the image.The eight-direction valley detection algorithm was used to extract the vein lines and the look-up table was proposed to thin the vein lines quickly.These improvements reduce the time of image processing greatly.(3)Taking into account the situation of finger translation and rotation,the template after the overlay is selected as the matching template,which improves the accuracy of matching.Aiming at the two modes of verification and identification,the two matching strategies,direct match and match after classification are designed which can keep the system in real time when the user volume is large.(4)To accelerate the development of software,the OpenCV library is ported for the embedded Linux platform,and the software function of the finger vein recognition system is realized by using the key library function.The experimental results show that the recognition rate of the miniaturized finger vein recognition system is up to 98.0%,and the equal error rate in the verification mode is 1.72%,and the response time is less than 1.5 seconds which verified that the system has high recognition accuracy and real-time performance.