Design And Implemention Of Video Analytics System Based On Edge Computing

As a typical use case of the Internet of Things,video analystics has been widely used in the fields of public safety,smart city and security.With the number of cameras increases and the camera generation is high-definition video streams,directly uploading video streams to the central cloud server for analytics will consume a large amount of transmission bandwidth and require large video analysis latency.Edge Computing can provide a platform for processing video data at the edge of the network.By offloading video data on edge nodes for analytics,the computational pressure and bandwidth load on the central cloud can be reduced,and the real-time performance of 'video analytics will be guaranteed..Video analytics is a computationally intensive application with high requirements on the computing resources of the server.Due to the limited computing power of edge computing,the design and implementation of video analytics system based on edge computing brings challenges.How to optimize the video analytics process of the system,so as to ensure the accuracy of the analytics and reduce the computing resource consumption on the edge server,becomes the key to realizing the real-time video analysis system.We design and implement the video analytics system on the lightweight edge computing platform.At the same time,through the optimization of the analysis process of the video analytics system to improve the real-time performance and the analytics results will interacts with users in real time.This paper mainly completes the following work:(1)Investigated the development background and research significance of video analytics system,the development status of edge computing.Based on the lightweight virtualization provided by container,a lightweight edge video analysis framework is proposed to analyze video data in real time and efficiently by migrating video analysis tasks to edge nodes.(2)The video analysis system is designed and implemented on the Lightweight Edge Server.The main functional modules of the system include video stream acquisition module,video analytics module,video storage function module and client display module.The video analysis functions include video-based face recognition,indoor positioning and speech recognition,and the visual web interface interacting with the users in real time is also designed.(3)Aiming at the specific application scenarios of the system,an optimization scheme of the video analytics process of the system is proposed.The optimization scheme includes video key frame filtering,video frame area compression and face database automatic updating.The purpose of optimization is to ensure the system analytics accuracy and improve the system running efficiency on the platform.In addition,this paper also performs functional verification and performance testing on the video analytics system.The performance test of the system includes memory occupancy,video analysis efficiency and analysis accuracy.At the same time,the performance of the system and the cloud-based application are compared to verify that the video analytics system has real-time characteristics.