基于深度学习的镜下岩石、矿物薄片识别

doi:10.13745/j.esf.sf.2023.6.7

地学前缘 ›› 2024, Vol. 31 ›› Issue (3): 498-510.DOI: 10.13745/j.esf.sf.2023.6.7

基于深度学习的镜下岩石、矿物薄片识别

张利军¹^,²^,³(), 鲁文豪², 张建东²^,^*(), 彭光雄², 卜建财¹, 唐凯¹, 谢渐成¹, 徐质彬¹, 杨海燕¹

1.湖南省遥感地质调查监测所, 湖南长沙 410015
2.中南大学有色金属成矿预测与地质环境监测教育部重点实验室, 湖南长沙 410083
3.湖南省自然资源事务中心洞庭湖区生态环境遥感监测湖南省重点实验室, 湖南长沙 410004

收稿日期:2022-12-30 修回日期:2023-05-14 出版日期:2024-05-25 发布日期:2024-05-25
通信作者: *张建东(1978—),男,博士,讲师,硕士生导师,矿物学、岩石学、矿床学专业。E-mail: csuzjd@sina.com
作者简介:张利军(1987—),男,硕士,高级工程师,主要从事资源遥感方向研究。E-mail: 275328308@qq.com
基金资助:
湖南省地质勘查项目“湖南省雪峰弧形成矿带转折段遥感信息提取及找矿靶区优选(20200806)”;湖南省重点研发计划项目“湖南省锂铌钽稀有金属资源高效勘查与开发(2019SK2261)”;湖南省地质院科研项目“基于全谱段波谱特征的岩矿精准识别技术研究(HNGSTP202315)”

Rock and mineral thin section identification based on deep learning

ZHANG Lijun¹^,²^,³(), LU Wenhao², ZHANG Jiandong²^,^*(), PENG Guangxiong², BU Jiancai¹, TANG Kai¹, XIE Jiancheng¹, XU Zhibin¹, YANG Haiyan¹

1. Hunan Remote Sensing Geological Survey and Monitoring Institute, Changsha 410015, China
2. Key Laboratory of Metallogenic Prediction of Nonferrous Metals and Geological Environment Monitor (Central South University), Ministry of Education, Changsha 410083, China
3. Hunan Key Laboratory of Remote Sensing Monitoring of Ecological Environment in Dongting Lake Area,Hunan Natural Resources Affairs Center, Changsha 410004, China

Received:2022-12-30 Revised:2023-05-14 Online:2024-05-25 Published:2024-05-25

摘要/Abstract

摘要：

岩石、矿物显微图像的识别是岩矿鉴定的基础手段之一,对地质资源勘探有着重要意义。薄片显微图像一般情况下是在实验室中进行的,这项工作繁琐费时,需要大量的人力资源,并且准确性受限于鉴定者的经验。深度学习智能图像识别算法可以通过卷积神经网络提取显微图像的深层特征,从而达到对显微图像进行快速、准确分类识别的目的。本研究以PyCharm平台为深度学习框架,以中国科学数据网上的南京大学教学岩石薄片数据集、南华北石炭纪灰岩显微图像数据集等6个数据集为基础制作了可以应用于岩石-矿物显微图像分类识别训练的数据集,搭建具有针对性的VGG卷积神经网络模型,该模型具有对整个岩石薄片图像与单个矿物图像分别提取其深层中的特征信息的能力,从而达到识别岩石薄片的目的。实验结果显示,随着模型训练迭代的进行,预测值与真实值之间的损失函数在不断减小,识别准确率在不断增加,在分别经过50个和30个循环训练之后,模型的损失函数与识别准确率已经基本收敛。模型对显微图像测试集的识别成功率均高于90%,说明搭建的模型对于图像有很好的特征提取效果,可以完成岩石-矿物显微图像识别的任务。通过本文的研究,可以认识到,深度学习对于处理岩矿鉴定这样的任务有着高超的效率与准确度,开发相关的模型并运用到前端软件上,可以加快矿产资源勘探工作的速度,对于生产实践有着重要的应用意义。

关键词: 岩矿鉴定, 深度学习, 卷积神经网络, 机器学习, 图像识别

Abstract:

Rock-mineral microscopic image identification is one of the basic means of rock and mineral identification, which is of great significance to the exploration of geological resources. Thin-section microscopic images are generally carried out in the laboratory. This work is tedious and time-consuming, requires a lot of human resources, and the accuracy is limited by the experience of the expert. Deep learning intelligent image recognition algorithm can extract the deep features of microscopic images by convolutional neural network, to achieve the purpose of fast and accurate classification and recognition of microscopic images. In this study, the PyCharm platform is used as the deep learning framework, and the data set that can be applied to the classification and recognition of rock-mineral microscopic images is made based on six data sets such as the teaching rock slice dataset of Nanjing University and the Carboniferous limestone microscopic image dataset of South North China on the China Science Data Network. We design a VGG convolutional neural network model. The model can analyze the feature information in the deep layer of the whole rock slice image and the single mineral image respectively, to achieve the purpose of identifying rock slices. The test results show that with the increase of model training times, the loss function between the predicted value and the real value is decreasing, and the recognition accuracy is increasing. After 50 and 30 cycles of training, the loss function and recognition accuracy of the model have been basically convergent. The recognition success rate of the model for the microscopic image test set is higher than 90%, indicating that the model has a good feature extraction effect for the image and can complete the task of rock-mineral microscopic image recognition. Through the research of this paper, it can be realized that deep learning has high efficiency and accuracy for dealing with such tasks as rock and mineral identification. Developing relevant models and applying them to front-end software can speed up the speed of mineral resources exploration and has important application significance for production practice.

Key words: rock identification, deep learning, convolutional neural network, machine learning, image recognition

中图分类号:

TP753
P585

张利军, 鲁文豪, 张建东, 彭光雄, 卜建财, 唐凯, 谢渐成, 徐质彬, 杨海燕. 基于深度学习的镜下岩石、矿物薄片识别[J]. 地学前缘, 2024, 31(3): 498-510.

ZHANG Lijun, LU Wenhao, ZHANG Jiandong, PENG Guangxiong, BU Jiancai, TANG Kai, XIE Jiancheng, XU Zhibin, YANG Haiyan. Rock and mineral thin section identification based on deep learning[J]. Earth Science Frontiers, 2024, 31(3): 498-510.

图/表 19

图1 研究技术路线示意图

Fig.1 Schematic diagram of research technical route

图2 岩石薄片显微图像数据集

Fig.2 Dataset of rock thin section microphotography

图3 矿物薄片显微图像数据集

Fig.3 Dataset of mineral thin section microphotography

表1 岩石薄片显微图像数据集组成

Table 1 Summary of rock thin section microphotography

类型	数量	单位
白云岩	155	张
灰岩	151	张
砂岩	332	张
总体	638	张

表2 矿物薄片显微图像数据集组成

Table 2 Summary of mineral thin section microphotography

类型	数量	单位
橄榄石	174	张
普通辉石	184	张
角闪石	106	张
黑云母	244	张
斜长石	112	张
红柱石	93	张
十字石	122	张
石榴子石	95	张
阳起石	34	张
鲕粒	103	张
总体	1 267	张

图4 翻转原始显微照片(a)、垂直翻转图显微照片(b)和水平翻转图显微照片(c)

Fig.4 Flip original microphotography (a), vertical flip microphotography (b), and horizontal flip microphotography (c)

图5 裁剪原始显微照片(a)与随机裁剪显微照片(b)

Fig.5 Clip original microphotography (a), and random clip flip microphotography (b)

图6 旋转原始显微照片(a)与旋转显微照片(b)

Fig.6 Rotating original microphotography (a), and rotating microphotography (b)

图7 VGG模型结构图

Fig.7 VGG model structure diagram

图8 岩石VGG卷积神经网络基本训练流程图

Fig.8 Basic training flow chart of VGG convolution neural network for rock

图9 矿物VGG卷积神经网络基本训练流程图

Fig.9 Basic training flow chart of VGG convolution neural network for mineral

图10 岩石显微薄片图像分类损失函数图

Fig.10 Loss function diagram for rock microphotography classification

表3 岩石训练集显微图像分类识别概率

Table 3 Recognition probability of microphotography classification in rock training set

训练集图像	分类识别概率/%
训练集图像	灰岩	砂岩	白云岩
灰岩	97.642	0.562	2.457
砂岩	1.125	98.214	0.661
白云岩	1.233	1.224	97.543

图11 岩石显微薄片图像分类精度变化图

Fig.11 Accuracy diagram for rock microphotography classification

图12 VGG11、VGG16和VGG19验证集精度变化对比图

Fig.12 Comparison diagram of accuracy in VGG11, VGG16, and VGG19 verification sets

图13 矿物薄片显微图像分类损失函数图

Fig.13 Loss function diagram of mineral microphotography classification

表4 矿物训练集显微图像分类识别概率

Table 4 Recognition probability of microphotography classification in mineral training set

训练集图像	分类识别概率/%
训练集图像	橄榄石	辉石	角闪石	黑云母	斜长石	红柱石	十字石	石榴子石	阳起石	鲕粒
橄榄石	98.863	0.31	0.128	0.357	0.275	0	0.382	0.532	0.457	0.028
辉石	0.321	98.61	0.41	0.146	0.246	0.532	0.253	0.248	0.254	0.148
角闪石	0.691	0.105	97.932	0.213	0.183	0.425	0.537	0.396	0.098	0.258
黑云母	0.241	0.261	0.025	98.254	0.261	0.216	0.429	0	0.248	0.427
斜长石	0.032	0.327	0.251	0.023	98.124	0.527	0.053	0.268	0.014	0.658
红柱石	0.124	0.141	0.521	0.245	0	96.534	0.293	0.421	0.054	0.320
十字石	0.231	0	0.326	0.612	0.342	0.623	97.632	0.134	0.147	0.247
石榴子石	0.052	0.056	0.153	0	0.279	0.495	0.327	97.894	0.014	0.089
阳起石	0.04	0.047	0.059	0.125	0.124	0.325	0.058	0.051	98.673	0.013
鲕石	0	0.143	0.195	0.025	0.166	0.323	0.036	0.056	0.041	93.812

图14 部分矿物显微薄片图像分类精度变化图

Fig.14 Accuracy diagram of mineral microphotography classification

图15 VGG11、VGG16和VGG19验证集精度变化对比图

Fig.15 Comparison diagram of accuracy in VGG11, VGG16, and VGG19 verification sets

参考文献 46

[1]	KERR P F, ROGERS A F. Optical mineralogy[M]. 4th ed. New York: McGraw-Hill, 1977.
[2]	常丽华, 曹林, 高福红. 火成岩鉴定手册[M]. 武汉: 地质出版社, 2009.
[3]	陈曼云, 金巍, 郑常青. 变质岩鉴定手册[M]. 北京: 地质出版社, 2009.
[4]	LAUNEAU P, CRUDEN A R, BOUCHEZ J L. Mineral recognition in digital images of rocks: a new approach using multichannel classification[J]. The Canadian Mineralogist, 1994, 32(4): 919-933.
[5]	付光明, 严加永, 张昆, 等. 岩性识别技术现状与进展[J]. 地球物理学进展, 2017, 32(1): 26-40.
[6]	李娟, 孙惠兰, 侯庆香. XRD 岩性快速识别方法研究[J]. 中国石油石化, 2017, 10: 61-62.
[7]	GOTTLIEB P, WILKIE G, SUTHERLAND D, et al. Using quantitative electron microscopy for process mineralogy applications[J]. Journal of the Minerals, Metals & Materials Society, 2000, 52(4): 24-25.
[8]	NIE J S, PENG W B. Automated SEM-EDS heavy mineral analysis reveals no provenance shift between glacial loess and interglacial paleosol on the Chinese Loess Plateau[J]. Aeolian Research, 2014, 13: 71-75.
[9]	TOVEY N K, KRINSLEY D H. Mineralogical mapping of scanning electron micrographs[J]. Sedimentary Geology, 1991, 75(1/2): 109-123.
[10]	BIAO F, GUO R X, JAMES C H, et al. Recognition and (semi-)quantitative analysis of REE-bearing minerals in coal using automated scanning electron microscopy[J]. International Journal of Coal Geology, 2024, 282: 104443.
[11]	FANDRICH R, YING G, BURROWS D, et al. Modern SEM-based mineral liberation analysis[J]. International Journal of Mineral Processing, 2007, 84(1-4): 310-320.
[12]	LIANG L, FENG Y S, BI L L, et al. Identification of hydrothermal alteration and mineralization in the Sancha magmatic Cu-Ni-Au sulfide deposit, NW China: implications for timing and genesis of mineralization[J]. Ore Geology Reviews, 2022, 143: 104770.
[13]	HRSTKA T, GOTTLIEB P, SKALA R, et al. Automated mineralogy and petrology-applications of TESCAN integrated mineral analyzer (TIMA)[J]. Journal of Geosciences, 2018, 63(1): 47-63.
[14]	陈倩, 宋文磊, 杨金昆, 等. 矿物自动定量分析系统的基本原理及其在岩矿研究中的应用: 以捷克泰思肯公司 TIMA 为例[J]. 矿床地质, 2021, 40(2): 345-368.
[15]	谢小敏, 李利, 袁秋云, 等. 应用 TIMA分析技术研究 Alum 页岩有机质和黄铁矿粒度分布及沉积环境特征[J]. 岩矿测试, 2021, 40(1): 50-60.
[16]	徐园园, 谢远云, 康春国, 等. 松花江早更新世水系演化: 来自 TIMA 矿物和地球化学的证据[J]. 地质科学, 2022, 57(1): 190-206.
[17]	HOAL K O, STAMMER J G, APPLEBY S K, et al. Research in quantitative mineralogy: examples from diverse applications[J]. Minerals Engineering, 2009, 22(4): 402-408.
[18]	PÉREZ-BARNUEVO L, PIRARD E, CASTROVIEJO R. Automated characterisation of intergrowth textures in mineral particles: a case study[J]. Minerals Engineering, 2013, 52(S1): 136-142.
[19]	ESRA BAŞTÜRKCÜ, CEYDA ŞAVRAN, A EKREM YÜCE, et al. Revealing the effects of mechanical attrition applied on Eskisehir-Beylikova REE ore utilizing MLA[J]. Minerals Engineering, 2022, 186: 107733.
[20]	朱紫怡, 周飞, 王瑀, 等. 基于机器学习的锆石成因分类研究[J]. 地学前缘, 2022, 29(5): 464-475. DOI
[21]	慎国强, 王玉梅, 张繁昌, 等. 基于人工智能的川东北三叠系杂卤石地震识别[J]. 地学前缘, 2021, 28(6): 155-161. DOI
[22]	THOMPSON S, FUETEN F, BOCKUS D. Mineral identification using artificial neural networks and the rotating polarizer stage[J]. Computers & Geosciences, 2001, 27(9): 1081-1089.
[23]	MAITRE J, BOUCHARD K, BÉDARD L. Mineral grains recognition using computer vision and machine learning[J]. Computers & Geosciences, 2019, 130: 84-93.
[24]	RUBO R A, DE CARVALHO C C, MICHELON M F, et al. Digital petrography: mineralogy and porosity identification using machine learning algorithms in petrographic thin section images[J]. Journal of Petroleum Science & Engineering, 2019, 183: 106382.
[25]	程国建, 李碧, 万晓龙, 等. 基于 SqueezeNet 卷积神经网络的岩石薄片图像分类研究[J]. 矿物岩石, 2021, 41(4): 94-101.
[26]	徐述腾, 周永章. 基于深度学习的镜下矿石矿物的智能识别实验研究[J]. 岩石学报, 2018, 34(11): 3244-3252.
[27]	ZENG X, XIAO Y C, JI X H, et al. Mineral identification based on deep learning that combines image and MOHS hardness[J]. Minerals, 2021, 11(5): 506.
[28]	张野, 李明超, 韩帅. 基于岩石图像深度学习的岩性自动识别与分类方法[J]. 岩石学报, 2018, 34(11): 3244-3252.
[29]	郝慧珍, 顾庆, 胡修棉. 基于机器学习的矿物智能识别方法研究进展与展望[J]. 地球科学, 2021, 46(9): 3091-3106.
[30]	ROSS B J, FUETEN F, YASHKIR D. Automatic mineral identification using genetic programming[J]. Machine Vision and Applications, 2001, 13(2): 61-69.
[31]	RUISANCHEZ I, POTOKAR P, ZUPAN J, et al. Classification of energy dispersion X-ray spectra of mineralogical samples by artificial neural networks[J]. Journal of Chemical Information and Computer Sciences, 1996, 36(2): 214-220.
[32]	ALIGHOLI S, KHAJAVI R, RAZMARA M. Automated mineral identification algorithm using optical properties of crystals[J]. Computers & Geosciences, 2015, 85: 175-183.
[33]	陈宗铭, 唐玄, 梁国栋, 等. 基于深度学习的页岩扫描电镜图像有机质孔隙识别与比较[J]. 地学前缘, 2023, 30(3): 208-220. DOI
[34]	徐圣嘉, 苏程, 朱孔阳, 等. 基于深度学习的岩石薄片矿物自动识别方法[J]. 浙江大学学报(理学版), 2022, 49(6): 743-752.
[35]	郭艳军, 周哲, 林贺洵, 等. 基于深度学习的智能矿物识别方法研究[J]. 地学前缘, 2020, 27(5): 39-47. DOI
[36]	SU C, XU S J, ZHU K Y, et al. Rock classification in petrographic thin section images based on concatenated convolutional neural networks[J]. Earth Science Informatics, 2020, 13(4): 1477-1484.
[37]	赖文, 蒋璟鑫, 邱检生, 等. 南京大学岩石教学薄片显微图像数据集[J]. 中国科学数据, 2020, 5(3): 21-33.
[38]	白林, 魏昕, 刘禹, 等. 基于VGG模型的岩石薄片图像识别[J]. 地质通报, 2019, 38(12): 2053-2058.
[39]	LI N, HAO H, GU Q, et al. A transfer learning method for automatic identification of sandstone microscopic images[J]. Computers & Geosciences, 2017, 103: 111-121.
[40]	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
[41]	谢涛. 基于深度学习的微细粒矿物识别研究[D]. 徐州: 中国矿业大学, 2020.
[42]	李彦冬, 郝宗波, 雷航. 卷积神经网络研究综述[J]. 计算机应用, 2016, 36(9): 2508-2515, 2565. DOI
[43]	TSUJI T, YAMAGUCHI H, ISHII T, et al. Mineral classification from quantitative X-ray maps using neural network: application to volcanic rocks[J]. Island Arc, 2010, 19(1): 105-119.
[44]	周永章, 王俊, 左仁广, 等. 地质领域机器学习、深度学习及实现语言[J]. 岩石学报, 2018, 34(11): 3173-3178.
[45]	LECUN Y, BENGIO Y, HINTON G. Deep learning[J]. Nature, 2015, 521(7553): 436-444.
[46]	IMAMVERDIYEV Y, SUKHOSTAT L. Lithological facies classification using deep convolutional neural network[J]. Journal of Petroleum Science and Engineering, 2019, 174: 216-228.

基于深度学习的镜下岩石、矿物薄片识别

Rock and mineral thin section identification based on deep learning

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 19

参考文献 46

相关文章 15

编辑推荐

Metrics

本文评价

[1]	董少群, 曾联波, 冀春秋, 张延兵, 郝静茹, 徐小童, 韩高松, 徐辉, 李海明, 李心琦. 超深层致密砂岩裂缝测井识别深度核方法[J]. 地学前缘, 2024, 31(5): 166-176.
[2]	周永章, 肖凡. 管窥人工智能与大数据地球科学研究新进展[J]. 地学前缘, 2024, 31(4): 1-6.
[3]	万成舟, 季晓慧, 杨眉, 何明跃, 张招崇, 曾姗, 王玉柱. 基于渐进多粒度训练深度学习的矿物图像识别[J]. 地学前缘, 2024, 31(4): 112-118.
[4]	张焕宝, 贺海洋, 杨仕教, 李亚林, 毕文军, 韩世礼, 郭钦鹏, 杜青. 基于机器学习的埃达克质岩构造背景判别研究[J]. 地学前缘, 2024, 31(4): 417-428.
[5]	刘洋, 李三忠, 钟世华, 郭广慧, 刘嘉情, 牛警徽, 薛梓萌, 周建平, 董昊, 索艳慧. 机器学习:海底矿产资源智能勘探的新途径[J]. 地学前缘, 2024, 31(3): 520-529.
[6]	苏恺明, 徐耀辉, 徐旺林, 张月巧, 白斌, 李阳, 严刚. 鄂尔多斯盆地延长组多油源贡献比例与分布规律:基于机器学习与可解释性研究[J]. 地学前缘, 2024, 31(3): 530-540.
[7]	陶士振, 吴义平, 陶小晚, 王晓波, 王青, 陈胜, 高建荣, 吴晓智, 刘申奥艺, 宋连腾, 陈荣, 李谦, 杨怡青, 陈悦, 陈秀艳, 陈燕燕, 齐雯. 氦气地质理论认识、资源勘查评价与全产业链一体化评价关键技术[J]. 地学前缘, 2024, 31(1): 351-367.
[8]	王子烨, 左仁广. 基于多源数据融合的喜马拉雅淡色花岗岩识别[J]. 地学前缘, 2023, 30(5): 216-226.
[9]	蒋果, 周可法, 王金林, 白泳, 孙国庆, 汪玮. 基于深度学习的花岗伟晶岩型锂铍矿物识别研究[J]. 地学前缘, 2023, 30(5): 185-196.
[10]	宋轩宇, 许民, 康世昌, 孙立平. 基于机器学习的冰冻圈典型流域水文过程模拟研究[J]. 地学前缘, 2023, 30(4): 451-469.
[11]	陈宗铭, 唐玄, 梁国栋, 关子珩. 基于深度学习的页岩扫描电镜图像有机质孔隙识别与比较[J]. 地学前缘, 2023, 30(3): 208-220.
[12]	朱紫怡, 周飞, 王瑀, 周统, 侯照亮, 邱昆峰. 基于机器学习的锆石成因分类研究[J]. 地学前缘, 2022, 29(5): 464-475.
[13]	胡义明, 陈腾, 罗序义, 唐超, 梁忠民. 基于机器学习模型的淮河流域中长期径流预报研究[J]. 地学前缘, 2022, 29(3): 284-291.
[14]	张振杰, 成秋明, 杨玠, 武国朋, 葛云钊. 机器学习与成矿预测:以闽西南铁多金属矿预测为例[J]. 地学前缘, 2021, 28(3): 221-235.
[15]	左仁广. 基于数据科学的矿产资源定量预测的理论与方法探索[J]. 地学前缘, 2021, 28(3): 49-55.