Abstract: In response to the challenges and issues in urban public safety systems such as low utilization of visual resources, identifying and retrieving complex surveillance images, processing large volumes of video data, functional changes, and tight coupling between hierarchical levels of the system architecture, a video middle platform architecture based on microservices is proposed, including the system’s architecture, task and resource scheduling algorithms, and video image retrieval algorithms. The research explores the “large middle platform, small front platform” linking mode and designs a multi-level video middle platform architecture composed of service layer, capability layer, and resource layer. Common shared application functionalities are decomposed and designed as standardized, fine-grained microservice components deployed in the service layer. The functionalities supporting software operation and processing of surveillance video data are designed as capability engines and middleware based on containerization technology, deployed in the capability layer. AI-based police big data analysis chains and algorithm libraries are designed and deployed in the resource layer. This forms a logically independent embedded video middle platform system. The video middle platform designed in this paper has been successfully applied in urban security systems in cities such as Fuzhou, achieving significant application results.

Key words: video middleware, microservices, scheduling engine, image analysis, AI algorithm library

摘要： 针对城市公共安全系统中存在视图资源利用率低、复杂监控图像的识别与检索难、大量视频数据处理难、功能变更难，体系架构层次间紧密耦合等问题与挑战，提出一种基于微服务的视频中台体系架构，包括系统的体系架构、任务与资源调度算法和视频图像检索算法。研究“大中台、小前台”的中台链接模式，设计由服务层、能力层和资源层构成的多级视频中台架构。将常用共性应用功能分解，设计为标准化细粒度的微服务组件，部署在服务层。将支撑软件运行和处理监控视频数据的功能，设计为基于容器化技术的能力引擎与中间件，部署在能力层。设计基于AI的警务大数据分析链路和算法库，部署在资源层。由此形成一个逻辑独立的嵌入式视频中台系统，所设计的视频中台成功应用于福州等市的城市安全系统，取得良好的应用成效。

关键词: 视频中台, 微服务, 调度引擎, 图像解析, AI算法库