基于高光谱图像的鞍山式铁矿品位反演方法研究

doi:10.3964/j.issn.1000-0593(2024)12-3494-10

摘要
参考文献
相关文章 (15)

全文: PDF (44886 KB)
输出: BibTeX | EndNote (RIS)

摘要：铁矿品位是评价铁矿贫富程度和经济价值的重要指标，铁矿品位的检定效率对铁矿开采效率具有重大影响。鉴于高光谱图像在物质分类与含量反演等领域具有分析速度快、准确性高、无破坏性等优势，分别在可见光-近红外（VIS-NIR）与短波红外（SWIR）两个波段范围内采集鞍山式铁矿的高光谱图像，探讨基于高光谱图像实现鞍山式铁矿品位反演的可行性。首先，提取高光谱图像感兴趣区（ROI）中的平均光谱代表对应样本的光谱数据，分别采用多元散射校正（MSC）与标准正态变量变换（SNV）对其进行光谱变换。然后，分别利用蒙特卡洛无信息变量消除（MCUVE）、竞争性自适应重加权采样（CARS）和连续投影算法（SPA）提取变换前后光谱数据的特征波段。最后，利用径向基函数神经网络（RBFNN）和极限学习机（ELM）分别建立鞍山式铁矿品位的定量反演模型。结果表明，在VIS-NIR范围内的光谱数据，经MSC变换后，利用CARS提取的特征波段建立的ELM品位反演模型效果最优（R²=0.90，RPD=3.02，RMSE=3.27，MAE=2.77）。将MSC-CARS-ELM模型应用于鞍山式铁矿样品的VIS-SWIR高光谱图像，能够生成像素级的铁矿品位分布图。该研究为快速、有效地实现鞍山式铁矿品位反演及可视化提供了一种新方法，在地质采矿领域具有重要的应用价值。

关键词：鞍山式铁矿；高光谱图像；特征波段提取；机器学习；可视化

Abstract：Iron ore grade is an important index used to evaluate the degree of wealth and economic value of iron ore, and the verification efficiency of iron ore grade greatly influences the efficiency of iron ore mining. Because of the advantages of hyperspectral images in the fields of substance classification and content inversion, such as fast analysis speed, high accuracy, and non-destructive, this study collected hyperspectral images of Anshan-type iron ore in the two bands of VIS-SWIR and NIR, respectively, and discussed the feasibility of realizing grade inversion of Anshan type iron ore based on hyperspectral images. First, the average spectral representation in the ROI of the hyperspectral image is extracted, and the spectral data of the corresponding samples are transformed by multivariate scattering correction (MSC) and Standard normal variate transformation (SNV), respectively. Then, Monte Carlo uninformative variable elimination (MCUVE), competitive adaptive reweighted sampling (CARS), and successive projections algorithm (SPA) were used to extract the characteristic bands of the spectral data before and after the transformation. Finally, the quantitative inversion model of Anshan type iron ore's iron grade is established using radial basis function neural network (RBFNN) and extreme learning machine (ELM). The results show that after the MSC transformation of spectral data in the VIS-SWIR range, the ELM grade inversion model established by using the feature bands extracted by CARS has the best effect (R²=0.90, RPD=3.02, RMSE=3.27, MAE=2.77). Applying the MSC-CARS-ELM model to the VIS-SWIR hyperspectral image of an Anshan-type iron ore sample can generate a pixel-level iron ore grade distribution map. This study provides a new method for realizing grade inversion and visualization of Anshan-type iron ore quickly and effectively, which has important application value in geology and mining.

Key words：Anshan type iron ore; Hyperspectral image; Feature band extraction; Machine learning; Visualization

收稿日期: 2023-10-19 修订日期: 2024-03-16

中图分类号:

TP79

基金资助: 国家自然科学基金项目（52074064）资助

通讯作者: 毛亚纯 E-mail: maoyachun@mail.neu.edu.cn

作者简介: 曹旺，1996年生，东北大学资源与土木工程学院博士研究生 e-mail: 2010418@stu.neu.edu.cn

引用本文:

曹旺，毛亚纯，文杰，丁瑞波，徐梦圆，付艳华. 基于高光谱图像的鞍山式铁矿品位反演方法研究[J]. 光谱学与光谱分析, 2024, 44(12): 3494-3503.
CAO Wang, MAO Ya-chun, WEN Jie, DING Rui-bo, XU Meng-yuan, FU Yan-hua. Study on Inversion Method of Anshan-Type Iron Ore Grade Based on Hyperspectral Image. SPECTROSCOPY AND SPECTRAL ANALYSIS, 2024, 44(12): 3494-3503.

链接本文:

https://www.gpxygpfx.com/CN/10.3964/j.issn.1000-0593(2024)12-3494-10 或 https://www.gpxygpfx.com/CN/Y2024/V44/I12/3494

[1] Zhao Y M, Feng C Y, Li D X. Acta Geologic Sinica (English Edition), 2014, 88(6): 1895.
[2] Northey S, Mohr S, Mudd G M, et al. Resources, Conservation and Recycling, 2014, 83: 190.
[3] Vieira M D M, Goedkoop M J, Storm P, et al. Environmental Science & Technology, 2012, 46: 12772.
[4] Crowson P. Resources Policy, 2012, 37: 59.
[5] Li X, Wen J, Fu Y, et al. Infrared Physics & Technology, 2022, 123: 104170.
[6] Chen G, Sun Z, Liu Y. Journal of Nanomaterials, 2016, 2016: 1683416.
[7] Pasquini C. Analytica Chimica Acta, 2018, 1026: 8.
[8] Choros K A, Job A T, Edgar M L, et al. Sensors, 2022, 22(7): 2687.
[9] Tuşa L, Kern M, Khodadadzadeh M, et al. Minerals Engineering, 2020, 146: 106150.
[10] Gallie E A, McArdle S, Rivard B, et al. International Journal of Remote Sensing, 2002, 23(11): 2229.
[11] Abdolmaleki M, Consens M, Esmaeili K. Remote Sensing, 2022, 14(24): 6386.
[12] Mishra G，Govil H，Srivastava P K. Quaternary Science Advances, 2021, 4: 100036.
[13] Galdames F J, Perez C A, Estévez P A, et al. Chemometrics and Intelligent Laboratory Systems, 2022, 224: 104538.
[14] Johnson C L, Browning D A, Pendock N E. Economic Geology, 2019, 114(8): 1481.
[15] Krupnik D, Khan S. Earth-Science Reviews, 2019, 198: 102952.
[16] Dalm M, Buxton M W N, van Ruitenbeek F J A. Minerals Engineering, 2017, 105: 10.
[17] Lobo A, Garcia E, Barroso G, et al. Remote Sensing, 2021, 13(16): 3258.
[18] Dalm M, Buxton M W N, van Ruitenbeek F J A. Mathematical Geosciences, 2019, 51: 849.
[19] Mao Y, Huang J, Cao W, et al. Infrared Physics & Technology, 2021, 115: 103717.
[20] Bachmann C M, Ainsworth T L, Fusina R A, et al. Manifold Coordinate Representations of Hyperspectral Imagery: Improvements in Algorithm Performance and Computational Efficiency. IEEE International Geoscience and Remote Sensing Symposium, 2010, 2010: 4244.
[21] ZHANG Jing-jing, ZHOU Xiao-yong, LIU Qi(张晶晶, 周晓勇, 刘奇). Acta Optica Sinica(光学学报), 2013, 33(11): 1128001.
[22] Kuo B C, Ho H H, Li C H. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(1): 317.
[23] Huang G B, Zhu Q Y, Siew C K. Extreme Learning Machine: A New Learning Scheme of Feedforward Neural Networks, 2004 IEEE International Joint Conference on Neural Networks, 2004, 2: 985.
[24] MAO Ya-chun, WEN Jian, FU Yan-hua, et al(毛亚纯，温健，付艳华，等). Spectroscopy and Spectral Analysis(光谱学与光谱分析), 2022, 42(1): 68.