Research On Surveillance Video Acquistion And Adaptive Playback System With Edge Computing

Surveillance video technology is extensively used in many applications for its intuitionistic and abundant information content,becoming the most efficient means in security.The surveillance video technology based on the traditional cloud computing architecture is limited by network transmission and the computing and storage capacity of cloud servers,which is insufficient to satisfy the processing needs of current surveillance video systems.Edge computing technology effectively extends the boundaries of cloud computing by offloading some computing and analysis tasks to the edge nodes,providing a feasible solution to meet the needs of surveillance video systems.This paper proposes a surveillance video system solution based on edge computing technology to optimize the acquisition phase and adaptive playback phase of surveillance video,which effectively improves the system performance.The main research work of this paper is as follows:In the surveillance video acquisition phase,the paper utilizes the cloud-edge collaboration to pre-process surveillance video data at the edge and extract surveillance video keyframes.The residual U-Net neural network applies the extracted video keyframes to detect salient objects to obtain valuable information.According to the currently available resources,use the asynchronous advantage actor-critic A3 C algorithm to optimize the residual U-Net tasks offloading decision between cloud and edge nodes,therefore with an average latency reduction of 61.06%,efficiently enhancing real-time performance.In the surveillance video adaptive playback phase,the system utilizes the MPEG-DASH protocol to transmit video streams and the deep Q-learning DQN algorithm to optimize the video chunk adaptive bitrate ABR decision algorithm on the grounds of information such as network bandwidth and player buffer.Thereby the quality of the downloaded video chunks bitrate is improved by 40.67%,providing users with a more distinct video playback picture.The paper also proposes a user experience quality QoE estimation model,which takes into account the factors affecting the user viewing experience during the surveillance video adaptive playback.It establishes a QoE evaluation model by mapping objective Qo S indicators to QoE indicators to evaluate the adaptive playback process.The system optimizes the bitrate decision algorithm based on the QoE evaluation results,which results in an average increase in QoE value of 13.6%,improves the user experience quality.Finally,the paper implements a surveillance video system based on edge computing technology,including the surveillance video acquisition and the adaptive playback phase.Next,the system is evaluated and experimented with in multiple dimensions using publicly available datasets through building an experimental evaluation environment.The system prototype demonstrates the feasibility and superiority of the proposed surveillance video system,which provides an advanced solution for surveillance video.