露天开采矿区要素遥感提取研究进展及展望
Research progress and prospect of remote sensing-based feature extraction of opencast mining areas
通讯作者: 李 瑜(1982-),女,硕士,正高级工程师,主要研究方向为遥感技术应用及期刊编辑。Email:gtzyygliyu@163.com。
责任编辑: 李瑜
收稿日期: 2022-04-14 修回日期: 2022-10-18
基金资助: |
|
Received: 2022-04-14 Revised: 2022-10-18
作者简介 About authors
张 仙(1992-),女,硕士,工程师,主要研究方向为遥感技术应用及期刊编辑。Email:
露天开采矿区要素遥感提取是矿业活动监测研究中的热门话题,但少有对相关研究的系统梳理和总结。为此,该文首先对露天开采矿区要素进行了界定,按要素种类将要素提取分为单要素提取和多要素提取,并简述了与一般地物提取和土地利用分类的区别; 其次,简要总结了目前相关研究的遥感数据来源与处理平台; 然后,将露天开采矿区要素遥感提取方法分为目视解译方法、基于传统特征的方法和深度学习方法3类,分别总结其研究现状,并分析了各方法的优缺点以及适用情况; 最后,对露天开采矿区要素遥感提取的未来研究方向进行了展望。文章认为有效地利用多源多时相数据、更强特征提取能力网络和复杂场景优化方法,进一步推动矿区要素智能化、精细化和鲁棒性提取是未来发展的趋势。研究结果可为露天开采矿区要素遥感提取的研究与应用提供参考。
关键词:
The remote sensing-based feature extraction of opencast mining areas is a hot topic in research on the monitoring of mining activities. However, there is a lack of systematic reviews and summaries of relevant studies. Therefore, this study first defined the features of an opencast mining area, divided the feature extraction into single- and multi-feature extractions according to feature types, and briefly described the differences between the feature extraction of opencast mining areas and general surface feature extraction and land use classification. Then, this study briefly summarized the sources and data processing platforms of remote sensing images available in relevant studies. Subsequently, this study divided the remote sensing-based methods for the feature extraction of opencast mining areas into three categories, namely visual interpretation, traditional feature-based approach, and deep learning. Then, it summarized the research status of these methods and analyzed their advantages, disadvantages, and applicability. Finally, this study proposed the future research direction of the remote sensing-based feature extraction of opencast mining areas, holding that the future developmental trend is to further promote the intelligent, fine-scale, and robust feature extraction of mining areas by effectively utilizing multi-source and multi-temporal data, networks with a stronger feature extraction capacity, and methods for the optimization of complex scenes. The results of this study can be used as a reference for the study and application of remote sensing-based feature extraction of opencast mining areas.
Keywords:
本文引用格式
张仙, 李伟, 陈理, 杨昭颖, 窦宝成, 李瑜, 陈昊旻.
ZHANG Xian, LI Wei, CHEN Li, YANG Zhaoying, DOU Baocheng, LI Yu, CHEN Haomin.
0 引言
然而,现有矿产资源相关信息更新滞后,尤其境外统计信息获取困难、准确度差,使传统的实地调查或数据统计方法受到很大限制。遥感技术可不受地域限制,及时、长期、准确地获取矿区情况,将遥感技术应用于露天开采矿区的监测,可直观揭示研究区域的矿产开采现状和生态环境的情况,是一种快速、有效的手段[4]。而露天开采矿区要素遥感信息提取是对露天开采矿区进行遥感监测的基础,因此露天开采矿区要素的遥感提取一直是研究热点。学者们针对露天采场[5⇓-7]、尾矿库[8⇓-10]、矿区道路[11]等矿区单要素以及多要素[12⇓-14]进行了遥感提取研究,使用的方法包括目视解译[15-16]、基于传统特征的方法[17⇓-19]和深度学习方法[20-21]等。
露天开采矿区要素遥感提取是矿业活动监测研究中的热门话题,虽已有大量研究,但少有对相关研究的系统总结。本文尝试对目前国内外露天开采矿区要素遥感提取研究进行梳理,总结常用的遥感数据源类型、处理平台和提取方法,以期进一步厘清露天开采矿区要素遥感提取方法的研究现状,探究当前面临的挑战及未来的发展方向,为新时期露天开采矿区要素遥感提取的研究与应用提供参考。
1 露天开采矿区要素的界定
目前研究中常见的露天开采矿区遥感提取要素类别总结如表1所示,具体包括露天采场、集水坑等采矿区要素,矿石堆、选矿场、洗矿场等中转场地要素,排土场、废石堆、尾矿库等尾矿区要素,选矿厂、冶炼厂等矿山建筑物,采矿沉陷、地裂缝、崩塌、滑坡、泥石流等地质灾害要素,以及道路、植被、水体、裸土等矿山环境要素。按提取要素的类别数量可分为单一要素提取和多要素提取。
表1 露天开采矿区要素类别
Tab.1
序号 | 一级类别 | 二级类别 |
---|---|---|
1 | 采矿区 | 露天采场、集水坑等 |
2 | 中转场地 | 矿石堆、选矿场、洗矿场等 |
3 | 尾矿区 | 排土场、废石堆、尾矿库等 |
4 | 矿山建筑物 | 选矿厂、冶炼厂等 |
5 | 地质灾害 | 采矿沉陷、地裂缝、崩塌、滑坡、泥石流等 |
6 | 矿山环境 | 道路、植被、水体、裸土等 |
露天开采矿区要素的遥感提取与一般的地物遥感提取或土地利用分类存在一定区别:
1)单一要素提取。一般遥感影像的单一地物提取,目标模式相对固定,有较统一且明显的特征,如运动场、飞机、建筑物、道路等; 而露天开采矿区要素内容更丰富、形态各不同,如尾矿库一般由坝体、尾砂和废水等要素组成,形状多变、尺度不一,空间纹理、颜色等特征千差万别[22],提取难度也有所提升。
2)多要素提取及土地利用分类。通常土地利用分类以《土地利用现状分类(GB/T 21010—2017)》[23]为标准,将采矿、采石场等地面生产用地,排土(石)和尾矿堆放用地统称为采矿用地; 而露天开采矿区多要素提取则需对其再进行细分类,且采矿场、排土场等要素特征十分相近,更给提取增加了难度。另外,土地利用分类一般提取整幅影像中的所有类别,而露天开采矿区多要素提取既可以是包括所有类别的全要素提取,也可以选择几种要素,与背景区分开。
2 遥感数据源类型及处理平台
2.1 遥感数据源类型
露天矿区要素遥感提取的遥感数据来源主要有卫星和无人机2种,以卫星数据为主导。已有研究中常用的卫星遥感影像包括几十米或十几米分辨率的中等空间分辨率遥感影像[9,19,24]和几米到亚米分辨率的高空间分辨率影像[21,25⇓-27]。前者以国外的Landsat系列、Sentinel系列、ASTER等和国内的HJ系列为代表; 后者包括国外的WorldView系列、SPOT系列、GeoEye系列,以及国产的高分系列(GF-1,GF-2,GF-6等)、资源系列(ZY1-02C,ZY-3等)以及BJ-2等类型。卫星影像的优点在于大范围、长时序、动态观测,多用于大面积的矿区要素提取和矿区变化监测; 但卫星光谱影像易受云雾的影响,对于具体时间、具体位置的影像质量不能保证。
无人机(unmanned aerial vehicles,UAV)影像具有高精度、高时效性的优势,空间分辨率可达厘米级,对露天矿区小型要素的提取有很大优势。近年来,随着消费级UAV的普及,基于UAV影像的研究也越来越多。如李鹏飞等[28]选取乌海市典型矿山排土场作为研究区,基于研究区UAV影像、选用8种可见光植被指数计算排土场坡面植被覆盖度,从而评价矿山排土场的植被恢复情况; 蔡祥等[20]以内蒙古某矿区为研究区,基于UAV影像、采用面向对象结合深度学习方法进行矿区地物的多要素提取,实验证明可将车辆、道路等小范围地物较为有效地提取出来; Xiang等[29]基于2014和2016年2 a 的多时相UAV影像,分别获取高空间分辨率数字高程模型(digital elevation model,DEM)数据,辅助进行露天开采矿区地貌信息的提取及变化分析。然而,UAV影像难以进行长时序的变化分析,而且其光谱信息取决于UAV搭载的传感器,受条件所限,目前多数研究中的UAV影像只有红、绿、蓝3个波段,光谱特征相比多光谱影像大大减少,这也制约了对部分露天矿区要素遥感提取的研究。
此外,雷达的应用可在地形高程上提供补充信息,能有效丰富要素特征。如卢遥等[30-31]和Nascimento等[32]分别利用激光雷达(light detection and ranging,LiDAR)数据得到地表高程信息,并结合GeoEye等高空间分辨率遥感影像进行矿区地物的协同提取,取得了较高的分类精度,也拓宽了数据类型范围; 方军[33]提出一种高效的点云分割方法,融合LiDAR点云与高分辨率遥感影像,对矿区建筑物的精细提取进行研究; 杨显华等[34]分别利用Stacking InSAR技术和人机交互解译方法对甘肃白银某煤矿区的Sentinel-1合成孔径雷达(synthetic aperture Radar,SAR)影像和亚米级光学遥感影像实施了采空塌陷识别和监测; 许凯等[35]基于Sentinel-1 SAR影像和DEM进行差分干涉雷达测量(differential interference SAR,DInSAR)得到干涉图和形变图,从而获取沉降信息特征,并根据沉降突变划分出可能为采矿区的区域,最后再结合高分辨遥感光学影像,利用深度卷积神经网络(deep convolutional neural networks,DCNN)方法识别水体、裸地和建筑物,进而得到目标区域的露天矿区分布信息图。
2.2 处理平台
传统的遥感数据处理平台多为单机式本地数据处理软件,如ENVI,ERDAS,eCognition和ArcMap等。随着卫星技术的迅速发展,卫星重访周期不断缩短,传感器空间分辨率不断提高,卫星遥感数据量急剧膨胀,目前已具备明显的大数据特征[36]: 大量化、多样化、快速化、价值密度低。传统处理平台已无法满足海量数据存储、高性能处理与分析、跨多平台分发等要求。
谷歌地球引擎(Google Earth Engine,GEE)是一个面向全球尺度的地理空间分析平台,充分集成了海量的地理和遥感数据资源以及强大的云端计算能力。它包括了多源遥感数据的管理、查询、可视化、下载、预处理和数据转换,并利用接口编程构建数据模型和后处理[37]。相比传统平台在计算机上存储、处理和分析,GEE平台可直接在云端进行,为大区域多时相遥感数据处理与分析提供了崭新的方向。
3 提取方法
本文根据露天开采矿区要素提取应用到的特征将提取方法分为目视解译方法、基于传统特征的方法和深度学习方法。
3.1 目视解译方法
目视解译得到的矿区要素边界平滑、类别全面、精度较高,迄今依然是矿区要素精确提取的主要方法之一。但其缺点在于需要作业人员有足够的专业知识,且耗费人力及时间,难以快速完成大范围的矿区要素提取,同时存在一定的主观性。
3.2 基于传统特征的方法
传统特征相对于深度特征而言为浅层特征,是可以进行可视化的特征,包括光谱特征、形状特征、统计特征、纹理特征和位置特征等。基于传统特征的方法主要有阈值法、支持向量机(support vector machine,SVM)、随机森林和最大似然分类等。这类方法发展时间长,比较成熟,是目前露天开采矿区多要素提取的主要方式,根据影像处理的最小基元,可分为基于像元和面向对象的提取2种。
3.2.1 基于像元的提取
除利用传统分类器进行多要素分类外,也有学者针对露天矿区设计新的特征来辅助要素提取。如Wu等[19]基于30 m空间分辨率的Landsat7 ETM+影像,从矿区形态特征入手,构建了基于像元的形态学矿区特征指数(morphological mining feature index,MMFI),然后采用阈值分割法提取了福建长汀县的露天矿区开采范围和时空变化,检测精度和总体精度均超过了0.85; Ma等[9]基于河北省宽城县长河矿区的Landsat8 OLI影像,通过分析铁矿区及尾矿池的光谱特征和结构熵,利用Landsat8 OLI第3,4,6,7波段构建了ULIOI指数,完成对铁矿区和尾矿池的提取; 朱彦光[13]对湖南省花垣县矿区的遥感影像构建了词包模型(bag-of-word,BOW),结合尺度不变特征变换(scale-invariant feature transform,SIFT)得到影像的特征直方图,并利用SVM方法完成蓄水池、废矿堆、矿山建筑、尾水、尾砂等露天矿区要素的提取。
综合上述研究可发现,基于像元的露天矿区多要素提取方法适用的数据多为30 m左右的中等空间分辨率影像,空间分辨率较粗,容易产生混合像元现象,因此更适用于提取露天矿区的整体范围; 对于矿区建筑、小型尾矿池等矿区内部要素,则难以精细提取。
3.2.2 面向对象的提取
随着传感器技术的发展,高空间分辨率影像越来越普及。高空间分辨率影像具有空间特征丰富的优点,但多数影像类型光谱分辨率相对较低,易产生“同物异谱”和“异物同谱”的问题。基于像元的方法无法充分利用影像的空间信息优势,浪费了较多的空间语义特征信息,而且易产生椒盐噪声[41]。在此情况下,面向对象的影像分析(object-based image analysis,OBIA)方法应运而生。面向对象的提取是先将影像分割成内部相对均一的斑块; 然后以斑块为单位,综合考虑其光谱、纹理、形状、空间关系等信息进行类别判定,充分发挥高空间分辨率影像空间特征的优势,且有效避免了椒盐现象,可以获得更高的分类精度。
目前,面向对象提取技术在露天矿区要素提取中应用较广泛。如范莹琳等[18]基于BJ-2的0.8 m分辨率影像,深度分析迁西地区铁尾矿的光谱、形状、纹理等特征,对铁尾矿进行提取,得到了较高的提取精度; 代晶晶等[6]选择基于边缘的分割算法进行影像分割,结合地形信息、光谱信息及几何信息建立规则集进行特征提取,最后采用隶属度函数法实现离子吸附型稀土开采区的提取; 刘雪龙[42]基于ASTER,SPOT-1和HJ-1A CCD影像的多源数据对磷石膏、锰渣等固体废物信息进行提取; 黄丹等[43]基于鄂尔多斯市某煤矿的SPOT-5影像,设计了4个层次,选择了光谱、距离、形状、纹理等十余个特征,多尺度地提取了采煤坑、露天煤矸石堆场、堆煤场及煤渣、矿区道路、建筑等11个类别。
影像分割是面向对象提取技术的核心之一,分割结果的精确度直接决定着提取精度的高低。常规影像分割方法包括ENVI软件中的多尺度分割法、eCognition软件中的分型网络演化算法(fractal network evolution algorithm,FNEA)、均值漂移算法、分水岭分割算法等。其中,前2种方法由于较大的便利性和较高的准确性,是目前研究中普遍应用的分割方法。但由于矿区地物类型多样,常规分割方法有时无法取得精确结果,因此发展新的分割方法也是目前研究重点之一。如彭燕等[45]提出一种视觉注意模型驱动的稀土矿区遥感信息智能提取方法,结合视觉注意模型和GrabCut算法自动分割影像,以提高稀土矿开采区的遥感识别精度。
面向对象提取技术有效发挥了高空间分辨率影像空间信息丰富的特点,很大程度上能够充分利用影像各种特征。但信息提取精度并不与特征类型的数量成正比,过多的特征(尤其是大量高相关性的特征)也会导致信息提取精度下降,即“维度灾难”。然而,目前露天矿区要素的面向对象提取研究中,较少见到对特征的筛选,而多是凭借经验选择或是对有限个类型特征组的比较,这不利于大量特征类型充分发挥作用。另外,受目标的形态、光照变化、背景等因素多样性的影响,无论是分割尺度还是特征的选择,都不具备较高的鲁棒性和可迁移性。
3.3 深度学习方法
基于深度学习的计算机视觉信息提取任务主要包括图像分类、目标检测和图像分割3类,其含义与基于传统特征方法的含义有所不同。计算机视觉任务中,图像分类是指仅判断检测图像中的类别,无需确定各类别的位置; 目标检测是框选出影像中目标位置并判定类别; 图像分割则是将图像中各类别按边缘区分开,并以不同颜色标注,并赋予类别名称,即对应基于传统特征方法中的“图像分类”。其中图像分割又可分为语义分割、实例分割和全景分割3种,语义分割对于同一类别的目标,无法区别不同个体; 实例分割结合了语义分割和目标检测技术,不仅区分类别,还可以区分同一类别的不同个体,但只面向图像中的目标; 而全景分割是在实例分割的基础上对图中的所有物体包括背景都进行检测和分割[48⇓-50]。由于相关技术和研究目的等原因,在目前的露天开采矿区要素遥感提取中,暂未见到实例分割和背景的相关研究。本文中基于深度学习的露天开采矿区遥感提取方法主要涉及计算机视觉领域的图像分类、目标检测和语义分割方法。
在目标检测方面,闫凯等[8]针对华北地区大型尾矿库目标,提出增加额外卷积层的策略改进了SSD模型结构,提高了对尾矿库的检测精度,精确率和召回率分别达到0.882和0.857。
在语义分割方面的要素提取方法较多,但以单一要素提取为主。如Gallwey等[24]基于加纳地区Sentinel-2影像,利用改进的U-Net模型进行了手工小规模采矿矿区范围提取,得到了2015—2018年当地手工采矿矿区的变化情况; 张成业等[21]基于GF-6影像,采用U-Net方法对滇东南个旧—马关都龙钨锡锑多金属基地典型矿区的尾矿库进行识别提取,并与传统机器学习方法对比,结果表明该方法在保证效率的同时取得了相对最高的提取精度; 张昆仑等[22]提出了一个多任务分支结构的分割网络(multi-task-branch network)完成了唐山尾矿库提取,在保证召回率为95.8%的情况下,尾矿库的检测准确率达到了78.8%。有学者通过各要素分别提取完成了露天开采矿区的多要素提取,如宋仁忠等[53]基于GF-2影像制作样本数据集,并采用U-Net模型分别完成了露天煤矿区中露天采场、矿区建筑物、堆煤场、道路、水体、裸地和植被的单类提取,获得了较高的精度。
另外,有部分研究涉及较为笼统的多要素提取,如蔡祥等[20]利用面向对象的影像分割算法制作数据标注标签,结合全卷积神经网络(fully convolutional networks,FCN)和U-Net网络集成的深度学习方法,对内蒙古西部某煤矿区进行分类,提取出矿区地面、道路、车辆和矿区建筑4类要素,并与传统面向对象方法和单一网络的深度学习方法做了对比,结果表明准确率和分类精度显著提高; 张峰极[54]基于GF-1,GF-2和Google Earth影像构建了铜陵市露天矿区的样本库,利用改进的DenseNet网络完成矿区、疑似矿区、水体、绿地、背景的五分类,结果证明露天开采矿区要素精度高于Deeplab,U-Net和Segnet网络模型。
在计算机视觉领域的图像分类方面有少量基于深度学习方法的露天开采矿区地物的多要素提取研究。如董畅[55]构建了基于卷积神经网络(convolutional neural network,CNN)的高分遥感影像多标签分类模型,结合CNN-RNN和注意力机制,完成了露天煤矿区的多标签分类,包括采场、中转场地、剥离区、排土场、建筑、草地、水域等12个类别,可较为有效地提取出草地、树木、建筑等要素; 但由于排土场和裸土以及水体与采场的颜色和纹理相似,且分布位置交错,所以这些类别的识别准确率较低。
4 结论与展望
4.1 结论
露天开采矿区要素遥感提取是矿业活动观测研究中的热门话题,本文对目前国内外的露天开采矿区要素遥感提取研究进行梳理,界定了露天开采矿区要素的范围,以及露天开采矿区要素的遥感提取与一般地物遥感提取和土地利用分类的区别,并简要总结了常用的遥感数据源类型与处理平台,在此基础上归纳了露天开采矿区要素遥感提取方法的类型和优缺点: 总的来说,目视解译类别全面、精度较高,但受人工效率所限应用范围小; 基于传统特征的提取方法中,相比基于像元的提取方法,面向对象的方法更适用于高空间分辨率影像,但受分割结果影响较大,且对尺度、特征等的定量化做的还不够; 深度学习方法无需人工选择特征,对单一要素提取的效果很好,但需要提前制作大量样本,且在多要素提取方面还有待加强。
4.2 展望
遥感数据的日益丰富与深度学习技术在遥感应用中的迅速发展使得矿区要素信息获取的智能化、完备性和精细化不断提高,有效地利用多源多时相数据、更强特征提取能力网络和复杂场景优化方法,进一步推动矿区要素智能化、精细化和鲁棒性提取是未来发展的趋势。
1)多源数据综合与高空间分辨率数据的应用。加强多光谱、高光谱、DEM,SAR,LiDAR等多源数据融合,提升矿区要素感知的总体信息量。
2)更新研究中的深度学习基础网络。现有研究基于的深度学习网络大多是2020年及以前的CNN,近年更强全局语义特征提取能力的Tranformer网络或Transformer与CNN混合网络已取得了精度上的大幅提升,矿区要素提取网络需要更新研究中的基础网络。
3)发展多要素提取的深度学习网络模型。深度学习方法对单一要素提取的准确性已得到广泛验证,然而由于矿区要素类型复杂的特点(如采矿区和尾矿区的特征高度相似),在多要素的提取方面还需进一步优化。未来可考虑针对要素特征加入特定层来优化网络模型等,有待学者们进一步研究。
4)发展适用于多类复杂要素的优化策略。多要素矿区模型优化的难点在于要素场景复杂与尺度多变,除了模型之外,优化策略也十分重要。当前相关研究讨论较少,一方面需要发展不同尺度要素均衡表达的损失函数,另一方面需要考虑有效的难例挖掘技术。
5)发展弱监督、非监督或域适应的识别技术。深度学习提取精度的重要影响因素之一是样本情况,尤其对于多要素提取来说,需要各要素类别的大量样本。因此,需发展弱监督、非监督或域适应的识别技术,降低对样本的依赖性,增强模型鲁棒性。
6)发展支持多源数据的多模态提取技术。现有研究大多基于单模态遥感数据,在现有技术下多已达到精度瓶颈,有效地利用多源数据需要发展多模态提取技术,对多源数据取长避短综合使用以进一步突破精度瓶颈。
参考文献
面向对象的离子吸附型稀土矿开采高分遥感影像识别方法
[J].
The object-oriented recognition method for remote sensing image with high spatial resolution for iron rare earth mining
[J].
Application of remote sensing,GIS and machine learning with geographically weighted regression in assessing the impact of hard coal mining on the natural environment
[J].
基于纹理的面向对象分类的稀土矿开采地信息提取
[J].
Extraction of rare earth mining areas using objects-oriented classification approach based on texture characteristics
[J].
基于改进DenseNet 网络的多源遥感影像露天开采区智能提取方法
[J].
Opencast mining area intelligent extraction method for multi-source remote sensing image based on improved DenseNet
[J].
基于面向对象分类的稀土开采区遥感信息提取方法研究
[J].
object-oriented classification for the extraction of remote sensing information in rare earth mining areas
[J].
Google Earth Engine平台支持下的铁矿区开采及植被变化遥感动态监测
[J].
Dynamic monitoring using remote sensing technology for iron mining area and vegetation change detection based on Google Earth Engine platform
[J].
基于深度学习的SSD模型尾矿库自动提取
[J].
DOI:10.7523/j.issn.2095-6134.2020.03.009
[本文引用: 2]
针对华北地区尾矿库自动提取问题,将基于深度学习的SSD目标检测模型应用于遥感图像尾矿库提取。首先标记华北地区2 000个样本,随机挑选1 500个作为训练样本,剩余样本作为测试样本,验证模型的检测精度。分析卷积层对应感受野与图像中尾矿库尺寸关系,发现原始SSD模型漏检误检大型尾矿库。改进SSD模型结构,提出增加额外卷积层的策略,提高对大型尾矿库目标的检测精度。实验表明,在置信度阈值为0.3时,改进的SSD模型相比原始模型,检测精确率提高10.0%,召回率提高14.4%,提高了大型尾矿库检测精度。验证了基于深度学习的SSD目标检测模型自动提取尾矿库的可行性以及改进算法的有效性。
Automatic extraction of tailing pond based on SSD of deep learning
[J]
DOI:10.7523/j.issn.2095-6134.2020.03.009
[本文引用: 2]
In order to automatically extract tailing ponds in North China, the SSD target detection model based on deep learning is applied to extract tailing ponds from remote sensing images. Firstly, 2 000 samples in North China were labeled as the foundation of database. 1 500 samples were randomly selected as training samples, and the remaining samples were used as test samples to verify the detection accuracy of the model. By using the original SSD model, the targets of large tailing ponds can not be detected accurately. In this work the relationship between the corresponding receptive field of convolution layer and the size of tailing reservoir in image was analyzed. Moreover, in order to improve the detection accuracy of the model for large-scale tailing pond targets, we modified the structure of SSD model by introducing an extra convolution layer. Experiments show that, compared with the original SSD model, the modified SSD model improves the detection accuracy by 10% and the recall rate by 14.4% at 0.3 confidence level. The detection accuracy for large tailing ponds is also improved. In this work, the feasibility of automatic extraction of tailing reservoir based on deep learning SSD model and the effectiveness of the modified algorithm were verified.
Remote sensing extraction method of tailings ponds in ultra-low-grade iron mining area based on spectral characteristics and texture entropy
[J].DOI:10.3390/e20050345 URL [本文引用: 3]
Extracting the tailings ponds from high spatial resolution remote sensing images by integrating a deep learning-based model
[J].
DOI:10.3390/rs13040743
URL
[本文引用: 1]
Due to a lack of data and practical models, few studies have extracted tailings pond margins in large areas. In addition, there is no public dataset of tailings ponds available for relevant research. This study proposed a new deep learning-based framework for extracting tailings pond margins from high spatial resolution (HSR) remote sensing images by combining You Only Look Once (YOLO) v4 and the random forest algorithm. At the same time, we created an open source tailings pond dataset based on HSR remote sensing images. Taking Tongling city as the study area, the proposed model can detect tailings pond locations with high accuracy and efficiency from a large HSR remote sensing image (precision = 99.6%, recall = 89.9%, mean average precision = 89.7%). An optimal random forest model and morphological processing were utilized to further extract accurate tailings pond margins from the target areas. The final map of the entire study area was obtained with high accuracy. Compared with the random forest algorithm, the total extraction time was reduced by nearly 99%. This study can be beneficial to mine monitoring and ecological environmental governance.
基于Canny 边缘检测算子的矿区道路提取
[J].
Extraction of roads in mining area based on Canny edge detection operator
[J].
模糊支持向量机和变化矢量分析相结合的矿区土地覆盖变化检测
[J].
Land cover change detection based on mixed dynamic monitoring method in mining area
[J].
DOI:10.13474/j.cnki.11-2246.2014.0355
[本文引用: 2]
针对目前土地覆盖变化检测常用的方法中存在有不同程度的误差累积,夸大了变化区域,本文提出了模糊支持向量机(FSVM)和变化矢量分析(CVA)相结合的土地覆盖检测方法。以某矿区2004年和2008年两期的CBERS遥感影像进行试验,结果表明,植被大幅减少,其他地类都有不同程度的增加,主要是由于开采规模和产量提升所致。通过与常规的其他两类方法比较发现,本文的方法的总体精度,Kappa系数,漏检误差和虚检误差分别为92.67%,0.8927,5.79%,7.31%,比其他两种方法有较大提高,能够提供较全面的变化类别和准确信息,可以有效应用于矿区土地覆盖动态监测。
基于尾矿库调查的西藏自治区金属矿开采强度分析
[J].
An analysis of mining intensity about metal mines based on investigation of tailing reservoirs in Tibet
[J].
遥感技术在南天山-昆仑山地区矿山开发占地调查中的应用
[J].
Application of remote sensing technology in the investigation of mine development land occupation in South Tianshan Kunlun Mountain area
[J].
Land use land cover change detection in the mining areas of V.D. Yalevsky coal mine-Russia
[C]//
基于面向对象的铁尾矿信息提取技术研究——以迁西地区北京二号遥感影像为例
[J].
Information extraction technologies of iron mine tailings based on object-oriented classification:A case study of Beijing-2 remote sensing images of the Qianxi area,Hebei Province
[J].
Detection of spatiotemporal changes of surface mining area in Changting Count Southeast China
[C]//
面向对象结合深度学习方法的矿区地物提取
[J].
Surface features extraction of mining area image based on object-oriented and deep-learning method
[J].
基于U-Net网络和GF-6影像的尾矿库空间范围识别
[J].
Recognition of the spatial scopes of tailing ponds based on U-Net and GF-6 images
[J].
基于MTBNet的唐山尾矿库提取
[J].
MTBNet for tailing pond of Tangshan City
[J].
A Sentinel-2 based multispectral convolutional neural network for detecting artisanal small-scale mining in Ghana:Applying deep learning to shallow mining
[J].DOI:10.1016/j.rse.2020.111970 URL [本文引用: 3]
基于WorldView II 图像的钨矿区水体信息提取方法研究——以江西大余县为例
[J].
A water information extraction method based on WorldView II remote sensing image in Tungsten ore districts:A case study of of Dayu County in Jiangxi Province
[J].
基于可见光植被指数的乌海市矿山排土场坡面植被覆盖信息提取研究
[J].
Vegetation coverage information extraction of mine dump slope in Wuhai City of Inner Mongolia based on visible vegetation index
[J].
Open-pit mine geomorphic changes analysis using multi-temporal UAV survey
[J].DOI:10.1007/s12665-018-7383-9 [本文引用: 1]
基于LiDAR与GeoEye的煤矿区典型地物协同提取
[C]//
Typical surface features extraction in mining area based on data of LiDAR and GeoEye
[C]//
综合机载LiDAR与高分影像的煤矿区典型地物提取方法
[J].
DOI:10.13474/j.cnki.11-2246.2015.378
[本文引用: 1]
综合利用LiDAR点云数据与WorldView-2高空间分辨率遥感影像,采用面向对象分类的矿区地表覆盖信息提取方法,利用nDSM高度阈值区分候选分割对象,构建了基于决策树分类器的矿区典型地物提取模型,在此基础上将图像光谱信息、DSM数据和地形参数等多源数据进行了融合,提取了出矸石堆、煤堆等典型煤矿区地物要素及植被、道路、水体等地表覆被要素信息。
Typical surface features extraction in mining area based on data of LiDAR and WorldView-2
[J].
DOI:10.13474/j.cnki.11-2246.2015.378
[本文引用: 1]
综合利用LiDAR点云数据与WorldView-2高空间分辨率遥感影像,采用面向对象分类的矿区地表覆盖信息提取方法,利用nDSM高度阈值区分候选分割对象,构建了基于决策树分类器的矿区典型地物提取模型,在此基础上将图像光谱信息、DSM数据和地形参数等多源数据进行了融合,提取了出矸石堆、煤堆等典型煤矿区地物要素及植被、道路、水体等地表覆被要素信息。
Land cover changes in open-cast mining complexes based on high-resolution remote sensing data
[J].
DOI:10.3390/rs12040611
URL
[本文引用: 1]
Remote sensing technologies can play a fundamental role in the environmental assessment of open-cast mining and the accurate quantification of mine land rehabilitation efforts. Here, we developed a systematic geographic object-based image analysis (GEOBIA) approach to map the amount of revegetated area and quantify the land use changes in open-cast mines in the Carajás region in the eastern Amazon, Brazil. Based on high-resolution satellite images from 2011 to 2015 from different sensors (GeoEye, WorldView-3 and IKONOS), we quantified forests, cangas (natural metalliferous savanna ecosystems), mine land, revegetated areas and water bodies. Based on the GEOBIA approach, threshold values were established to discriminate land cover classes using spectral bands, the normalized difference vegetation index (NDVI), normalized difference water index (NDWI) and a light detection and range sensor (LiDAR) digital terrain model and slope map. The overall accuracy was higher than 90%, and the kappa indices varied between 0.82 and 0.88. During the observation period, the mining complex expanded, which led to the conversion of canga and forest vegetation to mine land. At the same time, the amount of revegetated area increased. Thus, we conclude that our approach is capable of providing consistent information regarding land cover changes in mines, with a special focus on the amount of revegetation necessary to fulfill environmental liabilities.
基于多源遥感的采空塌陷识别技术应用研究
[J].
Application of mining collapse recognition technology based on multi-source remote sensing
[J].
基于DInSAR和DCNN的矿区分布信息提取方法及系统:中国,201811528248.X
[P]. 2019-04-02.
An extraction method and system of distribution information of mining area based on DInSAR and DCNN:China, 201811528248.X
[P]. 2019-04-02.
Google Earth Engine在土地覆被遥感信息提取中的研究进展
[J].
Application progress of Google Earth Engine in land use and land cover remote sensing information extraction
[J].
Google Earth Engine applications since inception:Usage,trends,and potential
[J].
DOI:10.3390/rs10101509
URL
[本文引用: 1]
The Google Earth Engine (GEE) portal provides enhanced opportunities for undertaking earth observation studies. Established towards the end of 2010, it provides access to satellite and other ancillary data, cloud computing, and algorithms for processing large amounts of data with relative ease. However, the uptake and usage of the opportunity remains varied and unclear. This study was undertaken to investigate the usage patterns of the Google Earth Engine platform and whether researchers in developing countries were making use of the opportunity. Analysis of published literature showed that a total of 300 journal papers were published between 2011 and June 2017 that used GEE in their research, spread across 158 journals. The highest number of papers were in the journal Remote Sensing, followed by Remote Sensing of Environment. There were also a number of papers in premium journals such as Nature and Science. The application areas were quite varied, ranging from forest and vegetation studies to medical fields such as malaria. Landsat was the most widely used dataset; it is the biggest component of the GEE data portal, with data from the first to the current Landsat series available for use and download. Examination of data also showed that the usage was dominated by institutions based in developed nations, with study sites mainly in developed nations. There were very few studies originating from institutions based in less developed nations and those that targeted less developed nations, particularly in the African continent.
Mapping mining areas in the Brazilian Amazon using MSI/Sentinel-2 imagery (2017)
[J].
DOI:10.3390/rs10081178
URL
[本文引用: 1]
Although mining plays an important role for the economy of the Amazon, little is known about its attributes such as area, type, scale, and current status as well as socio/environmental impacts. Therefore, we first propose a low time-consuming and high detection accuracy method for mapping the current mining areas within 13 regions of the Brazilian Amazon using Sentinel-2 images. Then, integrating the maps in a GIS (Geography Information System) environment, mining attributes for each region were further assessed with the aid of the DNPM (National Department for Mineral Production) database. Detection of the mining area was conducted in five main steps. (a) MSI (MultiSpectral Instrument)/Sentinel-2A (S2A) image selection; (b) definition of land-use classes and training samples; (c) supervised classification; (d) vector editing for quality control; and (e) validation with high-resolution RapidEye images (Kappa = 0.70). Mining areas derived from validated S2A classification totals 1084.7 km2 in the regions analyzed. Small-scale mining comprises up to 64% of total mining area detected comprises mostly gold (617.8 km2), followed by tin mining (73.0 km2). The remaining 36% is comprised by industrial mining such as iron (47.8), copper (55.5) and manganese (8.9 km2) in Carajás, bauxite in Trombetas (78.4) and Rio Capim (48.5 km2). Given recent events of mining impacts, the large extension of mining areas detected raises a concern regarding its socio-environmental impacts for the Amazonian ecosystems and for local communities.
基于最新国产卫星数据的尾矿库遥感监测
[J].
Remote Sensing monitoring of tailings ponds based on the latest domestic satellite data
[J].
基于SVM的大屯矿区遥感影像变化检测
[J].
Change detection of remote sensing images in Datun mining area based on support vector machine
[J].
Processes of land use change in mining regions
[J].
面向对象的煤矸石堆场SPOT-5 影像识别
[J].
DOI:10.3724/SP.J.1047.2015.00369
[本文引用: 1]
煤矸石是一种在成煤过程中与煤层伴生的黑灰色固体废弃物,不仅会污染环境,而且会严重损害附近居民的身体健康,目前已经成为矿区生态环境的主要影响源之一。因此,实时、准确、快速地获取煤矸石堆场的位置、形状和面积等信息,对于环境监测与管理具有重要的意义。本文以内蒙古鄂尔多斯市东胜区为试验区,将试验区内的典型地物分为:植被、水体、阴影、裸地、建筑、道路、排土排矸场、露天煤矸石堆场、堆煤场及煤渣、采煤坑和其他共11类。本文采用SPOT-5高分辨率遥感影像,面向对象提取研究区内的煤矸石堆场信息,并进行识别精度评价,精度达到89.47%。将面向对象的分类结果与最大似然分类方法的分类结果进行比较,结果表明,面向对象的提取方法可更好地应用于煤矸石堆场信息的自动提取,大幅度提高精度和效率。
Coal gangue yards information extraction using object-oriented method based on SPOT-5 remote sensing images
[J].
融合支持向量机和面向对象方法的矿区土地利用信息提取
[J].
Mining land use information extraction based on combining support vector machine and object oriented method
[J].
一种视觉注意模型驱动的稀土矿区遥感信息智能提取方法:中国, 201910317994.2
[P]. 2019-07-26.
A method of extraction of remote sensing information in rare earth mining areas driven by visual attention model:China,201910317994.2
[P]. 2019-07-26
A fast learning algorithm for deep belief nets
[J].
深度学习在图像识别中的应用研究综述
[J].
DOI:10.3778/j.issn.1002-8331.1903-0031
[本文引用: 1]
深度学习作为图像识别领域重要的技术手段,有着广阔的应用前景,开展图像识别技术研究对推动计算机视觉及人工智能的发展具有重要的理论价值和现实意义,文中对深度学习在图像识别中的应用给予综述。介绍了深度学习的由来,具体分析了深度信念网络、卷积神经网络、循环神经网络、生成式对抗网络以及胶囊网络等深度学习模型,对各个深度学习模型的改进型模型逐一对比分析。总结近年来深度学习在人脸识别、医学图像识别、遥感图像分类等图像识别应用领域取得的研究成果并探讨了已有研究值得商榷之处,对深度学习在图像识别领域中的发展趋势进行探讨,指出有效使用迁移学习技术识别小样本数据,使用非监督与半监督学习对图像进行识别,如何对视频图像进行有效识别以及强化模型的理论性等是该领域研究的进一步方向。
Survey of application of deep learning in image recognition
[J].
DOI:10.3778/j.issn.1002-8331.1903-0031
[本文引用: 1]
As an important technical means in the field of image recognition, deep learning has broad application prospects. Carrying out image recognition technology research has important theoretical and practical significance for promoting the development of computer vision and artificial intelligence. The application of deep learning in image recognition gives a review. The origin of deep learning is introduced. Deep learning models such as deep belief network, convolutional neural network, cyclic neural network, generated confrontation network and capsule network are analyzed. The improved models of each deep learning model are compared and analyzed one by one. In this paper, the research results of deep learning in image recognition applications such as face recognition, medical image recognition and remote sensing image classification are summarized. The existing researches are worth discussing. The development trend of deep learning in the field of image recognition is carried out. The discussion points out that the effective use of migration learning technology to identify small sample data, the use of unsupervised learning and semi-supervised learning to identify images, how to effectively identify video images and the theoretical significance of the model are further directions in this field.
基于深度学习的图像语义分割技术研究综述
[J].
DOI:10.3778/j.issn.1002-8331.1905-0325
[本文引用: 1]
图像语义分割技术是智能系统理解自然场景的关键技术之一,作为视觉智能领域的重要研究方向,该技术在移动机器人、无人机、智能驾驶以及智慧安防等领域具有广阔的应用前景。对于图像语义分割技术的研究发展历程进行了详细评述,包括从传统的语义分割方法到当前主流的基于深度学习的图像语义分割理论及其方法,重点阐述了基于深度学习的图像语义分割技术的框架及其实现过程,进而对当前具有代表性的典型算法的效果以及优缺点进行了分析,然后归纳了算法评价指标,最后对该技术的发展进行了总结与展望。该研究对于从事图像语义分割技术的研究人员和工程技术人员均具有很好的参考意义。
Survey of image semantic segmentation based on deep learning
[J].
DOI:10.3778/j.issn.1002-8331.1905-0325
[本文引用: 1]
Image semantic segmentation technology is one of the key technologies for intelligent systems to understand natural scenes. As an important research direction in the field of visual intelligence, this technology has broad application prospects in the fields of mobile robots, drones, intelligent driving and smart security. This paper gives a detailed review on the research and development of image semantic segmentation technology, including the traditional semantic segmentation method and the current mainstream image semantic segmentation theory based on deep learning, and the method of image semantic segmentation based on deep learning. It describes the framework and its implementation process, analyzes the effects, advantages and disadvantages of the typical representative algorithms, and then summarizes the algorithm evaluation indicators. Finally, the development of the technology is summarized and forecasted. The paper has a good reference for researchers and engineers who are engaged in image semantic segmentation technology.
基于深度学习的实例分割研究综述
[J].
A survey of instance segmentation research based on deep learning
[J].
Fully convolutional networks for semantic segmentation
[C]//
Region-based convolutional networks for accurate object detection and segmentation
[J].
DOI:10.1109/TPAMI.2015.2437384
PMID:26656583
[本文引用: 1]
Object detection performance, as measured on the canonical PASCAL VOC Challenge datasets, plateaued in the final years of the competition. The best-performing methods were complex ensemble systems that typically combined multiple low-level image features with high-level context. In this paper, we propose a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 50 percent relative to the previous best result on VOC 2012-achieving a mAP of 62.4 percent. Our approach combines two ideas: (1) one can apply high-capacity convolutional networks (CNNs) to bottom-up region proposals in order to localize and segment objects and (2) when labeled training data are scarce, supervised pre-training for an auxiliary task, followed by domain-specific fine-tuning, boosts performance significantly. Since we combine region proposals with CNNs, we call the resulting model an R-CNN or Region-based Convolutional Network. Source code for the complete system is available at http://www.cs.berkeley.edu/~rbg/rcnn.
基于改进UNet 孪生网络的遥感影像矿区变化检测
[J].
Remote sensing image mining area change detection based on improved UNet siamese network
[J].
基于深度学习和高分辨率遥感影像的露天矿地物分类方法
[J].
Classification of features in open-pit mining areas based on deep learning and high resolution remote sensing images
[J].
/
〈 |
|
〉 |
