Design And Implementation Of A Real-time Face Recognition System Based On Zynq

At present,various biometric technologies play an indispensable role in our lives.Face recognition is more widely used than other biometrics,whether it is attendance machines,unmanned ticket gates of train stations or Alipay.The face payment function is based on the prominent features of face recognition relative to other biometrics,such as fingerprint recognition and iris recognition,which are non-contact,non-invasive,and traceable.The development of traditional embedded products is separated by software and hardware design,and the update speed of hardware always fails to meet the requirements of software development,resulting in an increase in the iteration period of embedded products,which is abandoned by the market.In view of the shortcomings of traditional embedded development,this topic selects Zynq series fully programmable heterogeneous processor integrated with ARM processor and FPGA programmable logic resources to realize face recognition.In order to reduce the amount of calculation and meet the real-time face recognition,the face detection algorithm based on skin color and Viola-Jones detector is further researched.The face database is established by PCA dimension reduction technology.Based on the internal architecture of Zynq processor,combined with Vivado research,image preprocessing and IP core encapsulation are realized.The IP-based design concept is used to build the hardware engineering of face recognition system.For the complexity of embedded system development,PetaLinux tools are used to simplify the development of embedded systems.The system can be customized according to the designed hardware engineering,and the accompanying full-system simulator can be used to complete the simulation start of the system.The face recognition system designed in this paper is validated by ZedBoard platform,and the experimental results are analyzed.This project completed the design and verification of the face recognition system based on Zynq,and realized the software and hardware co-design.The experimental results show that the image preprocessing using the PL part of Zynq is about 3 times faster than the preprocessing in the PS part,and the recognition rate of the system is up to 82.5%.