摘要
相比于高光谱遥感和多光谱遥感,在无人机上搭载RGB相机,不仅操作简单,而且成本低廉。但是,目前基于无人机RGB相机进行矿区典型植物分类的研究较少。另外,开采沉陷对地表植物叶绿素含量的时空扰动规律尚不清楚。为解决上述问题,本研究融合RGB影像的光谱信息、纹理信息和点云的3D特征,使用神经网络、支持向量机、随机森林3种机器学习分类算法,实现了对采煤沉陷区典型植被的分类。基于多期影像的分类结果和植被指数,分析开采沉陷对典型植物叶绿素含量的时空扰动规律。研究表明,最佳的分类算法为支持向量机。多特征融合可以显著提高分类精度,相比于只用光谱特征,多特征融合后的总体分类精度提高了9.45%。总体分类精度可达90%,Kappa系数为0.906,可满足矿区植被调查的需要。通过分析针茅和柠条叶绿素含量的时空变化,发现采煤对拉伸区植被的影响最大,其次是压缩区和中性区。拉伸区应作为生态修复的重点区域。地裂缝是生态修复的重点对象。与针茅相比,柠条能更好地适应采煤引起的干扰,可作为生态恢复的先锋物种。
Compared with hyperspectral remote sensing and multispectral remote sensing,it is not only easy to operate,but also low cost to carry an RGB camera on the Unmanned Aerial Vehicle(UAV).However,there are few researches on the classification of typical plants in mining areas based on UAV-RGB cameras.In addition,the temporal and spatial disturbance of mining subsidence on chlorophyll content of surface plants is still unclear.In order to solve the above problems,this paper combines spectral information and texture information of RGB images and 3D characteristics of point cloud,and uses three machine learning classification algorithms,namely,neural network,support vector machine,and random forest,to achieve the classification of typical vegetation in mining subsidence areas.Based on the classification results of multi-period images and vegetation index(Blue-red ratio vegetation index,BRRI),the temporal and spatial disturbance regularity of mining subsidence on chlorophyll content of typical plants was analyzed.Research has shown that the best classification algorithm is support vector machine.Multi feature fusion can significantly improve classification accuracy,with an overall classification accuracy improvement of 9.45%compared to using only spectral features.The overall classification accuracy can reach 90%,and Kappa coefficient is 0.906,which can meet the needs of vegetation investigation in mining area.Analyzing the spatiotemporal changes in chlorophyll content of Stipa and Caragana,it was found that coal mining had the greatest impact on vegetation in the stretching zone,followed by the compression zone and neutral zone.The stretching zone should be regarded as the key area of ecological restoration.Ground fissure is the key object of ecological restoration.Compared with Stipa,Caragana can better adapt to the disturb-ance caused by coal mining and can be considered as a pioneer species for ecological restoration.
作者
陈凯
雷少刚
杨星晨
史运喜
陈树召
CHEN Kai;LEI Shaogang;YANG Xingchen;SHI Yunxi;CHEN Shuzhao(National Energy Group Baotou Energy Lijiahao Coal Mine,Ordos 017000,China;Ministry of Education Engineering Research Center for Mine Ecological Restoration,China University of Mining and Technology,Xuzhou 221116,China)
出处
《金属矿山》
CAS
北大核心
2023年第12期227-233,共7页
Metal Mine
基金
国家重点研发计划项目(编号:2016YFC0501107)
神华包头能源有限责任公司科技项目(编号:CEZB200204913)
鄂尔多斯科技合作重大专项(编号:2021EEDSCXQDFZ010)。
关键词
RGB
影像
矿区
机器学习
遥感分类
叶绿素含量
RGB image
mining area
machine learning
remote sensing classification
chlorophyll content
