基于实测光谱和国产高分五号高光谱卫星的铁尾矿表层含水率遥感反演方法研究

doi:10.3964/j.issn.1000-0593(2023)04-1225-09

摘要
参考文献
相关文章 (14)

全文: PDF (24471 KB)
输出: BibTeX | EndNote (RIS)

摘要：尾矿库作为高势能的人造泥石流危险源，在尾砂含水量过高时有溃坝风险，低含水量状态下产生扬尘则会危害周围环境。尾砂含水量实时、动态监测对于尾矿库安全状况及矿区环境保护具有重要意义。相比传统采样化验手段，高光谱遥感拥有观测面积大、数据实时易获取、光谱信息丰富的特点，为快速、高精度尾矿水分监测提供了手段。以鞍山-本溪铁矿群中的高硅型铁尾矿为研究区，实地采集尾砂样品77个，利用可见光-近红外（350～2 500 nm）光谱仪获取其光谱数据，分析不同含水率尾砂光谱特征及机理；引入竞争性自适应加权重采样法（CARS）筛选水分敏感波段，并基于敏感波段建立三维波段光谱指数（TBI），结合随机森林（RF）、粒子群优化的极限学习机（PSO-ELM）及卷积神经网络（CNN）算法建立尾砂水分反演模型，以国产高分五号高光谱卫星为数据源进行模型应用，获取尾矿库表层含水时空分布特征。结果表明：（1）尾砂光谱反射率随含水率升高明显下降，在1 455和1 930 nm处出现O—H吸收特征，吸收深度随含水率减小而逐渐减小；（2）基于CARS方法能够对高光谱数据（305波段）有效降维，筛选出18个水分敏感波段，进一步利用敏感波段构建不同形式的尾砂含水率TBI指数集，其中三维差值指数TBI5=(R_{1 097.47}-R_{1 990.67})-(R_{1 990.67}-R_437.39)，与水分含量相关性最高，达到0.84；（3）对比RF，PSO-ELM及CNN方法以及不同形式的输入数据，基于尾砂含水率TBI指数联合反射率数据集作为输入数据进行CNN建模，室内光谱模型达到验证精度R²=0.92，相对分析误差RPD=3.43，基于该模型利用高分五号卫星数据反演可获取研究区尾砂含水率空间分布结果，实地验证R²达到0.79，相对分析误差RPD=2.20，获得较好的预测效果。可为基于高光谱技术的铁尾矿水分含量大面积实时快速监测提供参考。

关键词：可见光-近红外光谱；三维光谱指数；国产高分五号卫星；铁尾矿；尾砂含水率

Abstract：The tailings dams, as a high-potential manufactured debris flow danger source, has the risk of dam failure when the moisture content is too high, and the generation of dust in the state of low moisture content will endanger the surrounding environment. Monitoring tailings’ moisture content is of great significance to the safety of tailings dams and the environmental protection of mining areas. Compared with traditional sampling and testing methods, hyperspectral remote sensing has the characteristics of a large observation area, easy acquisition of real-time data, and rich spectral information, which provides a means for rapid and high-precision monitoring of tailings moisture. The high-silicon type iron tailing in the Anshan-Benxi iron ore group was selected as the research area, and 77 tailings samples were collected on the spot.The spectral data was obtained by Vis-NIR (350~2 500 nm) spectrometer, and the competitive adaptive reweighted resampling method (CARS) was introduced to screen out the optimal bands and establish three-band spectral indices (TBI), combining random forest (RF), particle swarm optimization extreme learning machine algorithm (PSO-ELM) and convolutional neural network (CNN) model, a tailings moisture inversion model was established to obtain the spatiotemporal distribution characteristics of surface moisture in the tailings dam. Using the domestic Gaofen-5 hyperspectral satellite as the data source, the model was applied to obtain the temporal and spatial distribution characteristics of the surface moisturecontent in the tailings dam. The results showed that: (1) The spectral reflectance of tailings decreased significantly with the increase of moisture content, the spectra characteristic appeared in the O—H absorption bands at 1 455 and 1 930 nm, and the absorption depth gradually decreased with the decrease of moisture content; (2) Based on the CARS method, 18 moisture sensitive bands were screened out, and further use the sensitive bands to construct different forms of three-dimensional tailings moisture content characteristic spectral indices. It is proposed that TBI5=(R_{1 097.47}-R_{1 990.67})-(R_{1 990.67}-R_437.39), which has the highest correlation with moisture content, reaching 0.844 4; (3) Based on the three-dimensional spectral indices combined reflectance data set and the CNN method, the measured spectral model achieves the verification accuracy R²=0.92, the residual predictive deviation (RPD) =3.43. Based on this model, the spatial distribution results of tailings moisture content in the study area were obtained by inversion using the Gaofen-5 satellite data. The moisture content field verification model prediction result in R² reached 0.79.The result is relatively effective. This study can provide a reference for large-scale real-time, and rapid monitoring of iron tailings moisture content based on hyperspectral technology.

Key words：Visible-near-infrared spectra; Three-band spectral index; Domestic Gaofen-5 satellite; Iron tailings dam; Moisture content of tailings

收稿日期: 2022-02-08 修订日期: 2022-05-21

中图分类号:

TB79

基金资助: 国家自然科学基金项目（52074063，52074064），辽宁省科技计划项目（2020020307-JH1/103-02），辽宁省民生科技计划项目（2021JH2/10200035），国防重大项目2020GFZD002（N2001002）资助

通讯作者: 包妮沙 E-mail: baonisha@mail.neu.edu.cn

作者简介: 曹粤，1998年生，东北大学资源与土木工程学院硕士研究生 e-mail: caoyue.neu@foxmail.com

引用本文:

曹粤，包妮沙，周斌，顾晓薇，刘善军，虞茉莉. 基于实测光谱和国产高分五号高光谱卫星的铁尾矿表层含水率遥感反演方法研究[J]. 光谱学与光谱分析, 2023, 43(04): 1225-1233.
CAO Yue, BAO Ni-sha, ZHOU Bin, GU Xiao-wei, LIU Shan-jun, YU Mo-li. Research on Remote Sensing Inversion Method of Surface Moisture Content of Iron Tailings Based on Measured Spectra and Domestic Gaofen-5 Hyperspectral High-Resolution Satellites. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2023, 43(04): 1225-1233.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2023)04-1225-09 或 https://www.gpxygpfx.com/CN/Y2023/V43/I04/1225

[1] Cui Xiaowei, Geng Yong, Li Tao, et al. Resources, Conservation and Recycling, 2021, 173: 105746.
[2] YAN Sheng, YANG Jian-ying, SHI Chang-qing, et al(闫升，杨建英，史长青，等). Science of Soil and Water Conservation(中国水土保持科学), 2019，17(6)：111.
[3] MEI Guo-dong(梅国栋). Metal Mine(金属矿山), 2010，(10): 142.
[4] QIANG Jian-hua(强建华). Metal Mine(金属矿山), 2013，(10): 112.
[5] ZHU Yu-feng, LU Tie-ding, ZHOU Shi-jian, et al(朱煜峰, 鲁铁定, 周世健, 等). Metal Mine(金属矿山), 2012，(6): 92.
[6] Tian Jie, Zhang Baoqing, He Chansheng, et al. Agricultural and Forest Meteorology, 2019, 271: 225.
[7] YU Mo-li, LIU Shan-jun, SONG Liang, et al(虞茉莉, 刘善军, 宋亮, 等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2019, 39(10): 3096.
[8] LI Sheng-yang, LIU Zhi-wen, LIU Kang, et al(李盛阳, 刘志文, 刘康, 等). Infrared and Laser Engineering(红外与激光工程), 2019, 48(3): 9.
[9] LIU Zhao-xian, MENG Ya-bin(刘昭贤, 孟亚宾). Mine Surveying(矿山测量), 2019, 47(6): 43.
[10] WANG Chang-zuo, WANG Ji-hua, WANG Jin-di, et al(王昌佐, 王纪华, 王锦地, 等). Remote Sensing Information(遥感信息), 2003，(4): 33.
[11] David B L, Gregory P A. Soil Science Society of America Journal, 2002, 66(3): 722.
[12] YUE Sheng-ru, XU Zhen-xin, YANG Ming-hao, et al(岳胜如，徐振鑫，杨明豪，等). Hubei Agricultural Sciences(湖北农业科学), 2022，61(5)：181.
[13] Xing Z, Du C W, Shen Y Z, et al. Computers and Electronics in Agriculture, 2021, 191: 106549.
[14] Li H D, Liang Y Z, Xu Q S, et al. Analytica Chimica Acta, 2009, 648(1): 77.
[15] Morrisset D, Santamaria S, Hadden R, et al. Fire Safety Journal, 2022, 131: 103611.
[16] Wang Z, Zhang F, Zhang X L, et al. Science of The Total Environment, 2021, 775: 145807.
[17] Breiman L. Machine Learning, 2001, 45(1): 5.
[18] Nawar S, Mouazen A M. Soil and Tillage Research, 2019, 190: 120.
[19] WANG Hong, ZHANG Qiang, WANG Ying, et al(王宏, 张强, 王颖, 等). Computer Systems & Applications(计算机系统应用), 2021, 30(2): 201.
[20] Huang G B, Zhu Q Y, Siew C K. Neurocomputing, 2006, 70(1): 489.
[21] Moazen H, Molaei S, Farzinvash L, et al. Information Science, 2023, 628: 70.
[22] Navid K, Abidhan B, Pijush S, et al. International Journal of Thermal Sciences, 2022, 173: 107427.
[23] Wartini N, Budiman M, Maryam M, et al. Geoderma, 2019, 352: 251.
[24] Chen Y Y, Wang Z B. Chemometrics and Intelligent Laboratory Systems, 2018, 181: 1.
[25] Guo F, Xu Z, Ma H H, et al. Ecological Indicators, 2021, 133: 108400.
[26] Afriyie E, Verdoodt A, Mouazen A M. Soil and Tillage Research, 2022, 215: 105218.
[27] Sherman D M, Waite T D. American Mineralogist, 1985, 70: 1262.
[28] Li Q Q, Xie J C, Zhang J X, et al. Infrared Physics & Technology, 2020, 105: 103191.
[29] Evans James D. Straightforward Statistics for the Behavioral Sciences, Belmont, CA, US: Thomson Brooks/Cole Publishing Co, 1996.