联合深度学习与面向对象分析的卫宁北山露天矿山采场信息提取  

Open-pit mining area extraction based on deep learning and object-oriented image analysis in the Weining Beishan area

作　　者：刘丽 李士垚 王润 刘少宇 宋永飞 牛瑞卿[5] LIU Li;LI Shiyao;WANG Run;LIU Shaoyu;SONG Yongfei;NIU Ruiqing(Ningxia Survey and Monitor Institute of Land and Resources,Yinchuan 750002,China;Wuhan Center,China Geological Survey(Central South China Innovation Center for Geosciences),Wuhan 430205,China;Geological Environmental Center of Hubei Province,Wuhan 430034,China;Hubei Key Laboratory of Resources and Eco-environmental Geology,Wuhan 430034,China;School of Geophysics and Geomatics,China University of Geosciences,Wuhan 430074,China)

机构地区：[1]宁夏回族自治区国土资源调查监测院,宁夏银川750002 [2]中国地质调查局武汉地质调查中心(中南地质科技创新中心),湖北武汉430205 [3]湖北省地质环境总站,湖北武汉430034 [4]资源与生态环境地质湖北省重点实验室,湖北武汉430034 [5]中国地质大学地球物理与空间信息学院,湖北武汉430074

出　　处：《测绘通报》2024年第2期51-57,共7页Bulletin of Surveying and Mapping

基　　金：宁夏自然科学基金(2021AAC03432,2021AAC03431);宁夏回族自治区重点研发计划(2021BEG03001);中国地质调查局地质调查项目(DD20211391);资源与生态环境地质湖北省重点实验室(湖北省地质局)科技项目(KJ2023-18)。

摘　　要：卫宁北山地区是宁夏开展矿山生态环境恢复治理工作的关键节点区域。快速准确地获取区域内采场空间信息,监测矿山生态修复工程建设进展,已成为自治区矿政管理的重要工作之一。本文以卫宁北山地区为研究区,提出了一种联合深度学习与面向对象分析的国产高分辨率遥感卫星影像露天矿山采场信息提取方法。该方法首先采用支持小样本学习的U-Net模型进行露天采场的初步识别;然后结合面向对象分析与空间分析方法实现露天采场边界的精细化提取。经验证,该方法识别露天采场空间位置的精度为0.71,平均空间范围的提取精度为0.78。在此基础上,对卫宁北山地区露天矿山在2019—2021年的恢复治理情况开展识别与分析,识别出的125处露天矿山采场中有43.2%已开展生态修复工程,含44处采坑填埋与覆土整平、6处重新开发利用及4处人工复绿。结果表明,该方法在无须进行特征工程的前提下能够较为快速地对露天采场空间信息进行精细化提取,可为宁夏矿山遥感监测提供技术参考。Weining Beishan is a key area for the restoration and management of mining ecological environment in Ningxia.Quickly and accurately extracting information of the open-pit mining area and monitoring the progress of ecological restoration have become important tasks in the mining management.This article proposes a method for extracting information on open-pit mining areas from domestically produced high-resolution remote sensing satellite images in the Weinin Beishan area,which combines deep learning and object-oriented analysis.The method first uses the U-Net model,which supports small-sample learning,to perform initial recognition of the open-pit mining area,and then combines object-oriented analysis with spatial analysis methods to extract the mining area boundary.The experimental results show that the accuracy of identifying the spatial location of open-pit mining areas is 0.71,and the average spatial range extraction accuracy is 0.78.Based on this,the paper identifies and analyzes the ecological restoration status of open-pit mines in the Weining Beishan area from 2019 to 2021.Among the identified 125 open-pit mining areas,43.2%have been restored,including 44 filled and leveled pits,6 areas that have been redeveloped,and 4 areas that have been artificially revegetated.The method proposed in this article can accurately extract vector boundaries of open-pit mining areas without the need for feature engineering and can provide technical reference for mining remote sensing monitoring in Ningxia.

关 键 词：卫宁北山 露天矿山采场 深度学习 面向对象 遥感监测 

分 类 号：P237[天文地球—摄影测量与遥感]