枢纽级快件处理中心智能化视频监控与风险预警系统数据集

本数据集面向“监控系统响应时间≤1.5秒”的需求，数据来源于顺丰上海浦江中转场真实生产环境。2023.10-2024.3期间，部署枢纽级快件处理中心智能化视频监控与风险预警系统，经RTSP抽帧、Kafka缓存与Socket实时推送，同步嵌入YOLOv8+ByteTrack算法，对18类风险行为与9类安全事故进行识别-预警闭环测试，每30 ms记录一次端到端时延、GPU利用率、误报/漏报率，生成结构化数据。该数据集为系统功能以及性能定位、优化及响应指标验证提供可复现、可扩展的基准。数据量190MB。This dataset targets the requirement that the response time of the monitoring system shall be no more than 1.5 seconds. The data is collected from the real production environment of SF Express Shanghai Pujiang Transfer Station. From October 2023 to March 2024, an intelligent video monitoring and risk early warning system was deployed at the hub-level parcel processing center. The system integrated RTSP frame extraction, Kafka caching, Socket real-time push technologies and the YOLOv8+ByteTrack algorithm to conduct closed-loop recognition and early warning testing for 18 types of risky behaviors and 9 types of safety accidents. Structured data was generated by recording end-to-end latency, GPU utilization rate, false positive rate and false negative rate every 30 ms. This dataset provides a reproducible and scalable benchmark for system function verification, performance localization, optimization and response index validation. The total size of the dataset is 190 MB.

搜集汇总

背景与挑战

背景概述

该数据集基于顺丰上海浦江中转场真实环境，采集于2023年10月至2024年3月，通过集成YOLOv8与ByteTrack算法对18类风险行为和9类安全事故进行智能识别与预警测试，并记录系统性能指标。它为国家重点研发计划项目提供基准数据，用于验证和优化枢纽级快件处理中心的监控系统响应与风险预警能力。