Intelligent Video Analysis System for Smart Stores Based on 5G and Edge Computing
基于5G与边缘计算的智慧门店智能视频分析系统

摘要

在新零售背景下，传统门店监控存在视频传输延迟较高、带宽占用较大以及智能化程度较低等问题。为此围绕智 慧门店主题是围绕门店的监控+边缘AI算法，实现门店客流人数、性别、热力图等实时分析，为门店营销、数据分析、 复购等提供辅助决策，设计了一种基于5G及边缘计算的“端-边-云”三层架构解决方案。通过5G网络实现高清视频 的及时上传，在边缘侧进行本地化预处理与智能化分析，在云端对整体进行管理和深层次挖掘。实验结果表明，该系 统的端到端时延可以降到毫秒级别，带宽节省约40%，并且视频智能分析精度超过95%。同时从实时性、处理速度、 安全性以及成本等方面对该方案进行了论证，以期能给智慧门店监控建设起到一定作用。

Abstract

 In the context of new retail, traditional store monitoring systems face problems such as high video transmission latency, large bandwidth consumption, and a low level of intelligence. Focusing on the theme of smart stores, this paper centers on store monitoring and edge AI algorithms to realize real-time analysis of customer flow, gender distribution, heat maps, and other indicators, thereby providing decision-making support for store marketing, data analysis, customer repurchase, and related operations. A three-layer “device–edge–cloud” architecture solution based on 5G and edge computing is designed. High-definition video is uploaded in a timely manner through 5G networks, localized preprocessing and intelligent analysis are performed on the edge side, and overall management and in-depth data mining are carried out in the cloud. Experimental results show that the end-to-end latency of the system can be reduced to the millisecond level, bandwidth consumption can be reduced by approximately 40%, and the accuracy of intelligent video analysis exceeds 95%. The scheme is also discussed in terms of real-time performance, processing speed, security, and cost, with the aim of providing support for the construction of smart store monitoring systems.