矿山视频大数据智能分析与安全生产监控平台研究
Research on mine video big data intelligent analysis and work safety monitoring platform
蔡晨晖 1梁晓刚 1师剑雄 1白艳2
作者信息
- 1. 甘肃建投绿色建材产业发展集团有限公司,甘肃 兰州 730000
- 2. 中国科学院地理科学与资源研究所,北京 100101
- 折叠
摘要
为了实现对现代化大型矿山生产的全时段、全过程、全域覆盖的安全监测与监管,提出集成地理信息技术、5G技术、三维矿区及装备模型、遥感与视频监测、人工智能等多项信息技术的整体技术方案,并协同设计智慧矿山安全生产监控系统平台.研究结果表明:基于视频大数据智能分析技术,遴选视频目标自动检测算法,集成对YOLOv5 算法进行 Mosaic-9 数据增强、K-means聚类先验锚框与损失函数优化等改进技术,可实现现实场景中安全帽佩戴、反光衣穿戴以及烟雾火灾等视频自动化识别监测功能与虚拟三维模型的可视化融合,可为实现矿山安全全天候智能化、可视化监管提供分析平台.
Abstract
In order to realize the safety monitoring and supervision of modern large-scale mining production with whole time,whole process,and whole area coverage,an overall technical scheme integrating the geographic information technology,5G technology,3D mining area and equipment model,remote sensing and video monitoring,artificial intelligence and other infor-mation technologies was proposed,and the intelligent mine work safety monitoring system platform was designed collaborative-ly.The results show that based on the video big data intelligent analysis technology,the automatic detection algorithm of video target is selected,and the Mosaic-9 data augmentation,K-means clustering prior anchor box and loss function optimization of YOLOv5 algorithm are integrated,so as to realize the visual integration of video automatic recognition and monitoring functions such as helmet wearing,reflective clothing wearing and smoke fire in real scenes with virtual three-dimensional models,which can provide an analysis platform for realizing the all-weather intelligent visual supervision of mine safety.
关键词
目标检测/矿山安全监管/YOLOv5/三维可视化/监控系统Key words
object detection/mining safety supervision/YOLOv5/3D visualization/monitoring system引用本文复制引用
基金项目
兰州市科技计划项目(2021-1-20)
甘肃省科技计划项目(22CX8JA144)
出版年
2024
NETL
NSTL
维普
万方数据