Research And Implementation Of MEC Architecture For Video Analysis

With the rapid growth of video data,the demand for large-scale video processing tasks has increased dramatically.As one of the most popular smart services,video analysis plays an important role in many fields such as smart cities.Therefore,how to process video data in time to obtain effective information,and then provide users with video analysis services quickly is an important issue to be resolved.If the front-end computing resources are enriched and the front-end computing capacity is increased,the cost will be greatly increased.If the video analysis is directly offloaded to the remote cloud for processing,which will consume a lot of bandwidth,increase the burden on the core network,and cause a large video analysis delay.Mobile Edge Computing(MEC)can provide computing and storage services at the edge of the network.As the key technology for the evolution of the5 G network architecture,MEC is customizable and reconfigurable,which can meet the system's requirements for throughput,delay and intelligent deployment.Relying on MEC,various external applications can be provided at the edge of the network to make content and services close to users and provide a better user experience.At the same time,you can also choose the near-end MEC server to provide timely service support based on network conditions.A new deployment method of Edge Functions Modularized and Reorganized(EFMR)for large-scale video processing is proposed.This method sinks the video processing service to the edge of the network.Through the virtualization of network functions,the video service request sent to the edge server is divided into fine-grained functions according to the correlation of its inherent processing process,and resources are matched andredeployed on demand based on the division results.In this way,we can smoothly expand the edge video serviceprocessing capabilities at a small cost.Main tasks as follows:(1)The development background and research significance of video analysis based on deep learning technology are researched and explored,and the development and application scenarios of mobile edge computing are also investigated.Through container(Docker)virtualization technology,a two-layer edge video analysis framework is established,and video analysis tasks are processed quickly and efficiently at the edge nodes of the network.(2)Fine-grained decomposition of tasks and modular deployment of edge functions.The video processing service has obvious modular characteristics.A complete video processing service can often be broken down into multiple related modules,and different video analysis tasks often have the same processing part.In edge computing,virtualized technologies can also be used to pool computing,storage,and other resources.Based on business needs,it can provide distributed,low-latency,high-performance,secure,reliable,green,flexible,on-demand,and intelligent based on business needs.The energy-saving information infrastructure meets the needs of the business,so a modular processing model for video analysis is designed.(3)A video analysis system is designed and implemented on a lightweight edge server.The system is divided into a user layer and an edge-computing layer.In the edge-computing layer,a task decomposition module,a matching scheduling module,a control orchestration module,and an application module are also deployed.The video analysis system designed in this paper can realize target recognition and target detection based on deep learning technology.In the specific implementation,it is completed by the feature extraction module,SVM classification module and detection module,and each functional module is realized through Socket communication and data sharing.In addition,this article also performs function and performance tests on the built MEC-based video analysis system.The system performance tests include the evaluation of edge server computing volume,memory consumption.At the same time,this article compares the performance of the proposed method with traditional processing methods to verify that the video analysis system built in this article can meet the real-time requirements.