| In recent years,with the continuous improvement of software and hardware performance of various computing devices,artificial intelligence technology has begun to develop by leaps and bounds.As an important branch of artificial intelligence,computer vision has a very broad application prospect in social production.Among them,face recognition technology is one of the most deeply researched and widely used directions in computer vision,and is widely used in street surveillance,smart factories and other scenarios.The face recognition system is usually deployed in the cloud server,that is,after the camera captures the image,all the images are sent to the cloud server for data processing,analysis and storage and other operations.This form of processing can give full play to the computing and management advantages of cloud services,but at the same time,since the source data of face recognition are all video data,the data volume is large,and all uploading to the cloud server for processing will cause great computing and network pressure on the cloud..In order to reduce the pressure on the cloud server and improve the overall efficiency of the system,this thesis designs and implements a face recognition system based on edge-cloud collaboration,which can deploy algorithms on edge devices and realize edge-cloud collaboration in data and application management..The topic of this thesis comes from the enterprise scientific research cooperation project.The main work contents are as follows:(1)This thesis reviews the related technologies of face recognition system based on edge-cloud collaboration.Firstly,the current computer vision field,especially the various technologies in face recognition direction is reviewed.Secondly,the current edge-cloud collaboration technology and overall architecture are summarized.Finally,various development techniques applied in the system design and implementation are introduced.(2)This thesis designs a face recognition system based on edge-cloud collaboration.In this thesis,the overall architecture of the system is first designed and the cloud and edge parts of the system are layered according to functions;at the same time,a variety of communication processes are designed according to the actual needs of the system;Design and distribute it at both ends of the edge and cloud to realize data collaboration;at the same time,based on the characteristics of face recognition algorithm requiring task replacement,model upgrade,and code change,and combined with the design of the algorithm to achieve application management collaboration;Finally,the system is divided into modules,and describe the design of each module separately.(3)This thesis implements a face recognition system based on edge-cloud collaboration.In this thesis,the overall realization of the system is firstly planned,and the whole system is realized by modules,and the realization form and basic process of each module are described respectively.Finally,various functions of the system,system performance and algorithm performance are tested.,to ensure that the system achieves the design requirements in terms of overall function and performance. |