基于RGB数据的站台安全防控数据集

本数据集面向城市交通营运安全防控核心需求，针对现有数据集缺乏公共交通环境多样化、真实运行数据的痛点建设。数据来源于2024年11月广州市天河区石牌桥公交站北侧站台，基于实地模拟与监控拍摄结合的方式生成，为站台安全防控技术研究提供关键支撑。其产生方法具有明确针对性，由工作人员与学生共同模拟行人危险行为，分别模拟站台灭火器故障、不明遗留物品等场景，全面覆盖站台核心安全防控场景。数据集总计约37.23GB，包含18个类别，核心内容丰富：原始数据含2-90秒、30 帧/秒的视频及3000多张标注场景范围的图片；试验测试数据分为危险行为与异常行为相关训练测试文件；实验数据含3个权重文件；统计数据涵盖训练轮数、准确率等关键指标，配套6篇论文和4项专利。该数据集的构建填补了站台安全防控真实数据缺口，为人群危险行为检测、不明遗落物品检测及设施故障检测等研究提供高质量数据支撑，助力提升城市交通治理智能化水平。

This dataset is developed to meet the core requirements of urban traffic operation safety prevention and control, targeting the critical pain point that existing public transit datasets lack diverse and real-world operational data. The data was collected at the north platform of Shipaiqiao Bus Station, Tianhe District, Guangzhou in November 2024, and generated via a hybrid approach of on-site simulation and monitoring footage capture, providing pivotal support for technical research on bus platform safety prevention and control. Its generation method is purposefully targeted: staff and students jointly simulate dangerous pedestrian behaviors, including scenarios such as fire extinguisher malfunctions and unidentified unattended items on the bus platform, comprehensively covering all core safety prevention and control scenarios at the platform. The dataset has a total size of approximately 37.23 GB and consists of 18 categories, with rich core contents: the raw data includes videos with a duration of 2 to 90 seconds and a frame rate of 30 fps, as well as over 3,000 images annotated with scene scopes; the experimental test data is split into training and test files related to dangerous behaviors and abnormal behaviors; the experimental data includes 3 weight files; the statistical data covers key indicators such as training epochs and model accuracy, and is accompanied by 6 academic papers and 4 patents. The construction of this dataset fills the gap of real-world data for bus platform safety prevention and control, providing high-quality data support for researches such as crowd dangerous behavior detection, unidentified unattended item detection, and facility fault detection, and contributing to the improvement of the intelligent governance level of urban traffic.

搜集汇总

背景与挑战

背景概述

该数据集是针对城市交通站台安全防控需求构建的RGB视频和图像数据集，数据来源于2024年11月广州石牌桥公交站台，通过模拟行人危险行为（如灭火器故障、不明遗留物品）生成，覆盖核心安全场景。数据集包含约37.23GB的原始视频、标注图片、训练测试文件及实验数据，旨在为危险行为检测、设施故障检测等研究提供高质量真实数据支撑，以提升交通治理智能化水平。