面向煤矿综掘工作面复杂环境的视觉感知系统  

Visual perception system for complex environment of coal mine comprehensive excavation working face

作　　者：苏国用 胡坤[1,2,3] 王鹏彧[2,3] 赵东洋 张辉[3] SU Guoyong;HU Kun;WANG Pengyu;ZHAO Dongyang;ZHANG Hui(State Key Laboratory of Deep Coal Mining Response and Disaster Prevention and Control,Anhui University of Science and Technology,Huainan 232001,China;Collaborative Innovation Center for Mining Intelligent Technology and Equipment,Anhui University of Science and Technology,Huainan 232001,China;School of Mechatronics Engineering,Anhui University of Science and Technology,Huainan 232001,China)

机构地区：[1]安徽理工大学深部煤矿采动响应与灾害防控国家重点实验室,安徽淮南232001 [2]安徽理工大学矿山智能技术与装备省部共建协同创新中心,安徽淮南232001 [3]安徽理工大学机电工程学院,安徽淮南232001

出　　处：《浙江大学学报(工学版)》2025年第5期995-1006,1030,共13页Journal of Zhejiang University:Engineering Science

基　　金：安徽省高等学校科学研究资助项目(2022AH050834);国家自然科学基金:资助项目(52304166,52274153);深部煤矿采动响应与灾害防控国家重点实验室开放基金:资助项目(SKLMRDPC22KF24);安徽理工大学矿山智能技术与装备省部共建协同创新中心开放基金:资助项目(CICJMITE202206);安徽理工大学引进人才科研启动基金:资助项目(2022yjrc61)。

摘　　要：针对煤矿恶劣环境下视觉检测算法鲁棒性不足的难题,提出面向煤矿综掘工作面复杂环境的视觉感知系统.该系统采用ELAN-DS特征提取模块、SimAM注意力模块与解耦检测头对YOLOv7-tiny算法进行优化,构建煤矿综掘工作面视觉检测网络(CMCE-Net).将CMCE-Net迁移部署到视觉感知终端平台内,测试CMCE-Net在煤矿实际作业工况下的检测性能,基于煤矿综掘工作面数据集开展验证实验.实验结果表明,CMCE-Net的检测精度达到89.5%,相较于YOLOv7-tiny算法提升了5.2%.与Faster RCNN、YOLOv7-tiny、YOLOv8s等8种算法相比,综合检测性能最佳,模型复杂度处于较低水平.在视觉感知终端平台内,CMCE-Net对测试视频的检测速度最高达到33.4帧/s,在人机多目标混杂工况下,CMCE-Net对装备与人员的检测精度均大于90.0%.A visual perception system for the complex environment of the coal mine comprehensive excavation working face was proposed aiming at the problem of insufficient robustness of visual detection algorithms in the harsh environment of coal mines.ELAN-DS feature extraction module,SimAM attention module with decoupled detection head were used to optimize the YOLOv7-tiny algorithm in order to construct the coal mining comprehensive excavation face visual inspection network(CMCE-Net).CMCE-Net was migrated and deployed into the visual perception terminal platform in order to test the detection performance of CMCE-Net under the actual working conditions in coal mines.Validation experiments were conducted based on the data set of coal mine comprehensive excavation working face.The experimental results showed that the detection accuracy of CMCE-Net reached 89.5%,which was a 5.2%improvement compared with the YOLOv7-tiny algorithm.The combined detection performance was the best and the model complexity was at a lower level compared with eight algorithms such as Faster RCNN,YOLOv7-tiny and YOLOv8s.The detection speed of CMCE-Net on the test video reached up to 33.4 frames/s within the visual perception terminal platform,and the detection accuracy of CMCE-Net on the equipment and personnel was more than 90.0%under the human-machine multi-target mixing working condition.

关 键 词：综掘工作面 关键目标 视觉感知 检测网络 

分 类 号：TD76[矿业工程—矿井通风与安全] TP18[自动化与计算机技术—控制理论与控制工程]