基于视频分析的露天矿生产安全监测系统  

Open-pit Production Safety Monitoring System Based on Video Analysis

作　　者：王国昌 潘冰冰 李琦[1] WANG Guochang;PAN Bingbing;LI Qi(School of Information Engineering,Inner Mongolia University of Science and Technology,Baotou 014017,China)

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古包头014017

出　　处：《煤炭技术》2023年第11期162-166,共5页Coal Technology

基　　金：内蒙古关键技术攻关项目(2020GG0316);内蒙古自然基金资助项目(2020MS06008)。

摘　　要：为提高露天矿山自动管理能力,促进矿山生产安全高质量发展,利用深度学习和视频分析技术构建露天矿山生产安全监测系统。首先,设计矿山人工智能中台,统一管理适用于不同矿山作业场景的深度学习应用;其次,设计矿山实时视频分析推理告警机制,分析AI中台识别到的各种作业场景在时间和空间维度上是否合规,对违规场景进行告警;最后,告警机制联合控制摄像云台,对违规场景进行录像。结果表明:该系统对矿山作业现场实时视频流的分析速度达16.7 ms/帧,对违规场景的检出率92.3%,误检率3.6%,提高了露天矿山生产安全水平,为露天矿山的安全生产提供了良好的参考价值。In order to improve the automatic management capability of open-pit and promote the highquality development of open-pit production safety,an open-pit production safety monitoring system was developed using deep learning and video analytics.Firstly,an open-pit artificial intelligence(AI)midplatform was designed to manage deep learning applications of different open-pit working scenes.Secondly,an analysis and inference alarm mechanism was designed using open-pit real-time video to analyze whether the various working scenes identified by the AI mid-platform are compliant in timespace relation,and alarm the non-compliant scenes.Finally,non-compliant scenes will be recorded by the PTZ camera under control of the alarm mechanism.The results show that the system can analyze the real-time video stream of open-pit work scenes at a speed of 16.7 milliseconds per frame,with the detection rate of 92.3%and the missing rate of 3.6%for non-compliant scenes,improve the open-pit production safety level,and provide reference value for the production safety of open-pit.

关 键 词：生产安全 深度学习 视频分析 露天矿山 人工智能中台 云台控制 

分 类 号：X924.3[环境科学与工程—安全科学] TD76[矿业工程—矿井通风与安全]