TACSTechnology and Application of Computer Science2998-89262998-8934Art and Technology10.61369/TACS.2026060003Article基于 DeepSeek 的行为监测系统的设计与实现https://artdesignp.com/journal/TACS/3/6/10.61369/TACS.2026060003李晨,黄勇萍2026362026-03-28随着社会智能化和数字化的快速发展，公共安全、家庭安全和交通管理等领域对行为监测和异常情况检测的需求日益迫切。基于此，设计一个基于 DeepSeek 大模型与 YOLOv8n 融合的行为检测系统，实现视觉信号到自然语言描述的转换。系统采用分层架构设计，前端层基于微信小程序实现用户交互与视频展示，业务逻辑层通过云函数处理请求转发与用户认证，AI 服务层集成 YOLOv8n 模型进行实时目标检测，并通过规则引擎触发 DeepSeek API 调用，生成异常行为的自然语言描述。由于大模型 API 调用频率过高导致成本高的问题，设计了异常防抖与 LLM 冷却两级节流策略。行为监测, Deepseek, YOLOv8n, 异常检测

[1]Wang Q., Wang D., Lu J. et al.SAL-YOLO-DeepSeek:a lightweight real-time detection and LLM-driven decision framework for intelligent escalator safety monitoring[J]. Scientific Reports,2025,15(1):40600. [2]Chakurkar P, Kothari V.Towards safer environments: a YOLO and MediaPipe-based human fall detection system with alert automation[J].MethodsX,2025,13:103245. [3]Cao X, Li C, Zhai H.YOLO-AB:A Fusion Algorithm for the Elders' Falling and Smoking Behavior Detection Based on Improved YOLOv8[J].Computers, Materials & Continua,2025,83(3). [4]Shi H, Wang X., Shi J.Fall detection algorithm using enhanced HRNet combined with YOLO[J].Sensors,2025,25(13):4128. [5]Gong L., Yang H, Fang G., et al. A survey on video analytics in cloud-edge-terminal collaborative systems[EB/OL]. arXiv:2502.06581, 2025. https://doi.org/10.48550/arXiv.2502.06581.