基于无人机航测与改进深度卷积网络的尾矿库库容监测方法研究与应用  

Research and Application of Tailings Pond Storage Capacity Monitoring Method Based on Unmanned Aerial Vehicle Aerial Survey and Improved Convolutional Neural Network

作　　者：房衍志 杨修志 高世昌 张峥[1] 邹臣波 吴明阳 王梦晴 王昆[1] FANG Yanzhi;YANG Xiuzhi;GAO Shichang;ZHANG Zheng;ZOU Chenbo;WU Mingyang;WANG Mengqing;WANG Kun(School of Energy and Mining Engineering,Shandong University of Science and Technology;Mining Company of Shougang Group Co.,Ltd.)

机构地区：[1]山东科技大学能源与矿业工程学院,山东省济南市370100 [2]首钢集团有限公司矿业公司

出　　处：《现代矿业》2025年第2期186-189,共4页Modern Mining

摘　　要：为实现尾矿库库容的动态监测,提高风险防控能力,引入并改进DeepLabV3+卷积神经网络模型,使用无人机遥感技术获取库区地形数据,研究库容的智能化监测方法。以华东某尾矿库为例,结合像素配准和像元相加的方法,实现尾矿库面积与库容的精确计算。结果表明:改进DeepLabV3+模型的提取精度mIoU和mPA分别为97.89%和98.92%,满足工程监测要求,根据2次航测库容计算得出尾矿理论日堆存量为27 573 m^(3)。该监测方法自动化程度高、数据可靠,具备工程实践可行性和推广价值。To achieve dynamic monitoring of tailings pond storage capacity and enhance risk preven-tion and control capabilities,an improved DeepLabV3+convolutional neural network model is introduced,unmanned aerial vehicle remote sensing technology is used to obtain terrain data of the pond area,and an in-telligent monitoring method for pond capacity is studied.Taking a tailings pond in East China as an exam-ple,precise calculations of the pond area and capacity are achieved by combining pixel registration and pix-el addition methods.The results show that,by the improved DeepLabV3+model,the extraction accuracies of mIoU and mPA are 97.89%and 98.92%,respectively,which meet the requirements for engineering mon-itoring.The theoretical daily tailings discharge capacity calculated from two aerial surveys is 27573 m^(3).This monitoring method is highly automated,reliable in data,and has feasibility and promotion value in engi-neering practice.

关 键 词：尾矿库 无人机 卷积神经网络 库容监测 

分 类 号：TD926.4[矿业工程—选矿] TP18[自动化与计算机技术—控制理论与控制工程]