Identification of lithium-beryllium granitic pegmatites based on deep learning

doi:10.13745/j.esf.sf.2023.5.20

Abstract

Abstract:

Although remote sensing technology is widely used in large-scale exploration of metallic mineral resources, its application in direct rare-metal identification is limited, especially in the identification of hard rock Li/Be-bearing minerals. The problem is mainly due to low spectral resolution, low spatial resolution due to high physical similarity between ore body and wallrock, and small spectral difference between Li/Be-bearing minerals. To address this issue we investigate mineral identification methods based on deep learning models. Samples of Li-Be pegmatites and wallrock are collected from several pegmatite deposits and relevant spectral data are obtained. Spectral enhancement techniques are used to highlight the characteristic spectral features, and the characteristic absorption band similarity model and deep neural network models are compared for mineral identification accuracy. Results show that (1) the extracted characteristic absorption bands using a combination of envelope removal and mixed Gaussian model are better defined and reveal more geological insight. (2) Appropriate spectral enhancement can improve the accuracy of spectral models. In the case studied, the overall accuracy of the spectral model increases by 0.05 based on the logarithmic-first-order derivative spectrum over the original spectrum. (3) In terms of overall model accuracy, deep convolutional neural networks (0.78) perform better than shallow neural networks (0.55 for backpropagation; 0.73 for Extreme Learning Machines). Overall, the combination of hyperspectral imaging and deep convolutional neural network model can quickly and effectively identify pegmatite-hosted minerals, which offer a scientific basis for the direct identification of Li/Be-bearing minerals by satellite remote sensing.

Key words: lithium-beryllium, spectral transformation, MICA, deep learning, hyperspectral

CLC Number:

JIANG Guo, ZHOU Kefa, WANG Jinlin, BAI Yong, SUN Guoqing, WANG Wei. Identification of lithium-beryllium granitic pegmatites based on deep learning[J]. Earth Science Frontiers, 2023, 30(5): 185-196.

Figures/Tables 9

Fig.1 Training data example. (a) Spectral curve. (b) Spectral curve in binary form

Fig.2 Representative rock specimens. (a) Spodumene pegmatite; (b) lepidolite pegmatite; (c) Li-bearing tourmaline pegmatite; (d) beryl pegmatite.

Fig.3 Spectral model fit with absorption curve (left panel) and model fitting quality evaluation (right panel). (a1, a2) Spodumene; (b1, b2) Li-rich tourmaline; (c1, c2) lepidolite; (d1, d2) beryl.

Fig.4 Reflectance spectra of spodumene (black), Li-rich tourmaline (blue) and lepidolite (red) before (a) and after (b) spectral enhancement by envelope-removal technique

Fig.5 Reflectance spectra of beryl before (a) and after (b) envelop-removal enhancement

Fig.6 Characteristic absorption-band parameters for (a) spodumene, (b) lepidolite, (c) Li-rich tourmaline and (d) beryl

Fig.7 Spectral transformation results by different techniques

Table 1 Evaluation of ELM model accuracy under different spectral transformations

变换形式	神经元个数	训练精度	验证精度	测试精度
R	52	0.913	0.821	0.675
e^R	53	0.923	0.829	0.683
$R$	41	0.916	0.821	0.671
1/R	74	0.891	0.807	0.638
CR(R)	54	0.891	0.829	0.707
lnR	55	0.927	0.829	0.711
R'	57	0.886	0.843	0.719
R″	55	0.848	0.743	0.512
$1 R$ '	56	0.784	0.693	0.598
(lnR)'	78	0.927	0.857	0.728
(e^R)'	55	0.895	0.843	0.715
( $R$ )'	42	0.881	0.836	0.711
$1 R$ ″	56	0.843	0.721	0.451
(lnR)″	53	0.845	0.779	0.61
(e^R)″	54	0.857	0.743	0.541
( $R$ )″	64	0.861	0.736	0.533

Table 1 Evaluation of ELM model accuracy under different spectral transformations

变换形式	神经元个数	训练精度	验证精度	测试精度
R	52	0.913	0.821	0.675
e^R	53	0.923	0.829	0.683
$R$	41	0.916	0.821	0.671
1/R	74	0.891	0.807	0.638
CR(R)	54	0.891	0.829	0.707
lnR	55	0.927	0.829	0.711
R'	57	0.886	0.843	0.719
R″	55	0.848	0.743	0.512
$1 R$ '	56	0.784	0.693	0.598
(lnR)'	78	0.927	0.857	0.728
(e^R)'	55	0.895	0.843	0.715
( $R$ )'	42	0.881	0.836	0.711
$1 R$ ″	56	0.843	0.721	0.451
(lnR)″	53	0.845	0.779	0.61
(e^R)″	54	0.857	0.743	0.541
( $R$ )″	64	0.861	0.736	0.533

Table 2 Comparison of overall identification accuracy of different methods

方法	训练集精度	验证集精度	测试集精度
方法	训练集精度	验证集精度	R²	Kappa系数
MICA	0.719	0.637	0.496	0.431
BP	0.774	0.693	0.549	0.415
ELM	0.927	0.857	0.728	0.651
DCNN	0.966	0.882	0.776	0.716

References 48

[1]	王登红, 王成辉, 孙艳, 等. 我国锂铍钽矿床调查研究进展及相关问题简述[J]. 中国地质调查, 2017, 4(5): 1-8.
[2]	王登红, 孙艳, 刘喜方, 等. 锂能源金属矿产深部探测技术方法与找矿方向[J]. 中国地质调查, 2018, 5(1): 1-9.
[3]	王登红, 王瑞江, 孙艳, 等. 我国三稀(稀有稀土稀散)矿产资源调查研究成果综述[J]. 地球学报, 2016, 37(5): 569-580.
[4]	PAN Z W, LIU J J, MA L Q, et al. Research on hyperspectral identification of altered minerals in yemaquan west gold field, Xinjiang[J]. Sustainability, 2019, 11(2): 428. DOI URL
[5]	王瑞军, 李名松, 汪冰, 等. 新疆红山铜金矿床基于地面高光谱遥感找矿模型构建[J]. 现代地质, 2016, 30(3): 577-586.
[6]	CARDOSO-FERNANDES J, LIAM A, RODA-ROBLES E, et al. Constraints and potentials of remote sensing data/techniques applied to lithium (Li) pegmatites[J]. The Canadian Mineralogist, 2019, 57(5): 723-725. DOI URL
[7]	潘蒙, 唐屹, 肖瑞卿, 等. 甲基卡新3号超大型锂矿脉找矿方法[J]. 四川地质学报, 2016, 36(3): 422-425, 430.
[8]	CARDOSO-FERNANDES J, TEODORO A, LIMA A. Remote sensing data in lithium (Li) exploration: a new approach for the detection of Li-bearing pegmatites[J]. International Journal of Applied Earth Observation and Geoinformation 2019, 76: 10-25. DOI URL
[9]	金谋顺, 高永宝, 李侃, 等. 伟晶岩型稀有金属矿的遥感找矿方法: 以西昆仑大红柳滩地区为例[J]. 西北地质, 2019, 52(4): 22-231.
[10]	代晶晶, 王登红, 令天宇. 基于地面反射波谱技术的锂含量定量反演研究[J]. 遥感技术与应用, 2019, 34(5): 992-997.
[11]	代晶晶, 王登红, 代鸿章, 等. 川西甲基卡锂矿基地典型岩石及矿物反射波谱特征研究[J]. 岩矿测试, 2018, 37(5): 507-517.
[12]	HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7): 1527-1554. DOI PMID
[13]	Wu X W, SAHOO D, HOI S C H. Recent advances in deep learning for object detection[J]. Neurocomputing, 2020, 396: 39-64. DOI URL
[14]	ALSHAZLY H, LINSE C, BARTH E, et al. Ensembles of deep learning models and transfer learning for ear recognition[J]. Sensors, 2019, 19(19): 4139. DOI URL
[15]	刘彦锋, 张文彪, 段太忠, 等. 深度学习油气藏地质建模研究进展[J]. 地质科技通报, 2021(4): 235-241.
[16]	左仁广, 彭勇, 李童, 等. 基于深度学习的地质找矿大数据挖掘与集成的挑战[J]. 地球科学, 2021, 46(1): 350-358.
[17]	易敏, 叶发旺, 张川, 等. 基于深度学习的蚀变矿物识别: 以新疆白杨河铀矿床为例[J]. 铀矿地质, 2021, 37(4): 673-682.
[18]	陈睿华, 孙媛, 尚天浩, 等. 基于光谱变换的宁夏银北地区可溶性阴离子反演[J]. 中国土壤与肥料, 2022(8): 94-103.
[19]	朱赟, 申广荣, 项巧巧, 等. 基于不同光谱变换的土壤盐含量光谱特征分析[J]. 土壤通报, 2017, 48(3): 560-568.
[20]	郑煜, 常庆瑞, 王婷婷, 等. 基于连续统去除和偏最小二乘回归的油菜SPAD高光谱估算[J]. 西北农林科技大学学报(自然科学版), 2019, 47(8): 37-45.
[21]	李新澳, 甘淑, 胡琳, 等. 面向桉树和车桑子精准识别的高光谱特征变换方法比较分析[J]. 测绘科学技术学报, 2021, 38(4): 380-383.
[22]	贺军亮, 崔军丽, 李仁杰. 土壤重金属铬的高光谱估算模型[J]. 实验室研究与探索, 2019, 38(7): 8-11, 61.
[23]	CHEN C Q, JIANG Q G, ZHANG Z C, et al. Hyperspectral inversion of petroleum hydrocarbon contents in soil based on continuum removal and wavelet packet decomposition[J]. Sustainability, 2020, 12(10): 4218. DOI URL
[24]	LIU X Y, ZHANG J L, YIN Y F, et al. Underwater polarization image restoration based on logarithmic transformation and dark channel[J]. Optoelectronics Letters, 2020, 16(2): 149-153. DOI
[25]	CHOE E, MEER F V D, RUITENBEEK F V, et al. Mapping of heavy metal pollution in stream sediments using combined geochemistry, field spectroscopy, and hyperspectral remote sensing: a case study of the Rodalquilar mining area, SE Spain[J]. Remote Sensing of Environment, 2008, 112(7): 3222-3233. DOI URL
[26]	栾孟杰, 崔国范, 孙威. 基于幂次变换和高频加强滤波的乳腺图像增强算法[J]. 渤海大学学报(自然科学版), 2019, 40(4): 378-384.
[27]	孙林, 程丽娟. 植被叶片生化组分的光谱响应特征分析[J]. 光谱学与光谱分析, 2010, 30(11): 3031-3035.
[28]	王长耀, 牛铮, 唐华俊, 等. 对地观测技术与精细农业[M]. 北京: 科学出版社, 2001.
[29]	LIU H L, DING X Q. Handwritten character recognition using semi-tied full covariance Gaussian mixture model[J]. Chinese Journal of Electronics, 2005, 14(4): 649-652.
[30]	李昌利, 沈玉利. 期望最大算法及其应用(1)[J]. 计算机工程与应用, 2008, 44(29): 61-64.
[31]	KOKALY R F. PRISM: processing routines in IDL for spectroscopic measurements (installation manual and user’s guide, version 1.0)[R/OL]. U.S. Geological Survey Open-File Report, 2011: 432[2022-11-20].https://pubs.usgs.gov/of/2011/1155/.
[32]	HUANG G B, ZHU Q Y, SIEW C K. Extreme learning machine: theory and applications[J]. Neurocomputing, 2006, 70(1/2/3): 489-501. DOI URL
[33]	周飞燕, 金林鹏, 董军. 卷积神经网络研究综述[J]. 计算机学报, 2017, 40(6): 1229-1251.
[34]	余凯, 贾磊, 陈雨强, 等. 深度学习的昨天、今天和明天[J]. 计算机研究与发展, 2013, 50(9): 1799-1804.
[35]	GU J, WANG Z, KUEN J, et al. Recent Advances in Convolutional Neural Networks[J]. Pattern Recognition, 2015, 77: 354-377 DOI URL
[36]	MISHRA P K, SATAPATHY S C, ROUT M. Segmentation of MRI brain tumor image using optimization based Deep Convolutional Neural Networks (DCNN)[J]. Open Computer Science, 2021, 11(1): 380-390. DOI URL
[37]	宋光慧. 基于迁移学习与深度卷积特征的图像标注方法研究[D]. 杭州: 浙江大学, 2017.
[38]	ZHANG S W, ZHANG S B, ZHANG C L. Cucumber leaf disease identification with global pooling dilated convolutional neural network[J]. Computers and Electronics in Agriculture, 2019, 162: 422-430. DOI URL
[39]	YU Y, ZHANG K L, YANG L. Fruit detection for strawberry harvesting robot in non-structural environment based on Mask-RCNN[J]. Computers and Electronics in Agriculture, 2019, 163: 104846. DOI URL
[40]	黄双萍, 孙超, 齐龙, 等. 基于深度卷积神经网络的水稻穗瘟病检测方法[J]. 农业工程学报, 2017, 33(20): 169-176.
[41]	PAN S J, YANG Q. A survey on transfer learning[J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(10): 1345-1359. DOI URL
[42]	庄福振, 罗平, 何清, 等. 迁移学习研究进展[J]. 软件学报, 2015, 26(1): 26-39.
[43]	ZHANG Z P, DING J L, ZHU C M. Combination of efficient signal pre-processing and optimal band combination algorithm to predict soil organic matter through visible and near-infrared spectra[J]. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 2020, 240: 118553. DOI URL
[44]	CLARK R N, KING T V V, KLEJWA M, et al. High spectral resolution reflectance spectroscopy of minerals[J]. Journal of Geophysical Research: Solid Earth, 1990, 95(B8): 12653-12680.
[45]	WU M J, ZHOU K F, WANG Q, et al. Mapping hydrothermal zoning pattern of porphyry Cu deposit using absorption feature parameters calculated from ASTER data[J]. Remote Sensing, 2019, 11(14): 1729. DOI URL
[46]	HONG Y S, CHEN Y Y, YU L, et al. Combining fractional order derivative and spectral variable selection for organic matter estimation of homogeneous soil samples by VIS-NIR spectroscopy[J]. Remote Sensing, 2018, 10(3): 479. DOI URL
[47]	QIAO X X, WANG C, FENG M C, et al. Hyperspectral estimation of soil organic matter based on different spectral preprocessing techniques[J]. Spectroscopy Letters, 2017, 50(3): 156-163. DOI URL
[48]	GHOLIZADEH A, BORǓVKA L, SABERIOON M M, et al. Comparing different data preprocessing methods for monitoring soil heavy metals based on soil spectral features[J]. Soil and Water Research, 2015, 10(4): 218-227. DOI URL