Research On Mask-covered Face Recognition Technology And Application Based On IoT Securit

In recent years,face recognition technology has been popularized in all aspects of human life,and has been widely used in the field of safety,autonomous services,and fast online payment.However,main techniques of commonly used face recognition is based on the standard face without occlusion.When the face is occluded by the mask,the accuracy and efficiency of recognition become worse.Now,during the global outbreak of the COVID-19,people need to wear masks for safety protection when they are out.Therefore,in order to solve the potential safety hazards of taking the mask off in face recognition places,it is necessary to design a facial-recognition system for masks based on embedded devices of the Io T.This paper focuses face recognition system for mask occlusion.The main contents of this paper are classified into the following points:1)In the absence of an open source of mask face datasets,a manually generated virtual datasets was finally adopted by analyzing the collection method according to requirements.Based on the standard face datasets Celeb A,the mask face datasets were worn on fixed areas by combining Dilb face key point detection,face alignment,image stitching technology and face pose estimation,mask templates.2)According to the problem of face completion,the GAN that repairs missing images through the network generation method is selected for research.The BEGAN is selected as algorithm for face completion through compare experiment of the GAN and the improved GAN.Then,BEGAN combine with the context encoder to repair face occlusion by the mask.In order to improve algorithm completion effect,input of the generator of the BEGAN use the average face to replace the occluded of face.The performance of the improved face completion algorithm is promoted through compare face recognition experiment.3)Considering mask face recognition is applied to embedded devices,the architecture of the whole face recognition system is designed,mainly by the c + + language and QT interface design framework.First,the performance of face recognition model in server and embedded is tested.Then the overall framework of the Io T is designed,and the hardware equipment of the embedded development board is selected.Meanwhile,the audio and video compression technology and the transmission protocol of the Io T are compared,and the protocol in accordance with this paper is selected to complete the push and pull flow functions of the video.Finally,QT interface of the client is designed to complete the testing of video surveillance and face recognition function of masks.