Design And Implementation Of Iris Recognition System Based On Zynq

Today,information security has become an increasingly important research direction.Biometric recognition technology is an important means of protecting information security.Among many biometric recognition technologies,iris recognition technology has advantages in accuracy,reliability,security,etc.On the carrier of the iris recognition system,the PC has a large volume and is not suitable for mobile use.Although the embedded processor is small,it cannot meet the requirements of high-speed image processing due to the limitations of the serial computing architecture.In recent years,with the introduction of the ARM+FPGA heterogeneous computing platform and the application of software and hardware collaborative design ideas,it has become possible to conduct rapid image processing research work on embedded devices with the help of the heterogeneous computing platform.It is the main research goal of this paper to implement hardware acceleration of the iris recognition process,to achieve a small volume and good real-time iris recognition system.In this paper,a variety of iris algorithms are studied.Daugman extracts features based on 2D-Gabor wavelet coding,uses the iris recognition algorithm with Hamming distance matching,and its algorithm structure is suitable for hardware implementation;Boles uses the method of wavelet zero crossing when extracting and encoding iris features,which is sensitive to iris rotation and has poor robustness;Lim applies Haar wavelet to the feature vector extraction step,which has low resolution for feature extraction;By comparing the above algorithms,based on the iris recognition architecture proposed by Daugman and the heterogeneous system of ARM+FPGA,an iris recognition system is realized.The light source noise removal,histogram equalization,iris positioning,and feature comparison algorithm acceleration module are completed on the FPGA.Operating system transplantation,identification process control,DMA transfer,memory access,etc.are completed on ARM.In the research process,MATLAB,C++and Vivado were used for simulation.The system was tested on the Zynq platform,and the false acceptance rate(FAR)and false rejection rate(FRR)were counted.The performance was compared with the ARM platform and the recognition time was reduced from 1463ms to 997ms.Compared with the PC platform In comparison,the volume also meets the requirements for mobile use.