基于岩石图像深度学习的多尺度岩性识别

马泽栋; 马雷; 李科; 姚伟; 王培丁; 王鑫宇

doi:10.19509/j.cnki.dzkq.2022.0140

基于岩石图像深度学习的多尺度岩性识别

doi: 10.19509/j.cnki.dzkq.2022.0140

合肥工业大学资源与环境工程学院, 合肥 230009

基金项目:

国家自然科学基金项目 41831289

国家自然科学基金项目 42072276

大学生创新创业训练计划项目 202010359061

详细信息

作者简介:
马泽栋(2000-), 男, 主要从事水文地质方向的研究工作。E-mail: 1044929468@qq.com

通讯作者:
马雷(1985-), 男, 副教授, 主要从事裂隙水运移多尺度模拟方面的研究工作。E-mail: Lei8505@hfut.edu.cn

中图分类号: P585.2
计量
- 文章访问数: 1491
- PDF下载量: 217
- 被引次数: 0
出版历程
- 收稿日期: 2021-06-28

Multi-scale lithology recognition based on deep learning of rock images

School of Resources and Environmental Engineering, Hefei University of Technology, Hefei 230009, China

摘要

摘要:
岩性识别作为人工智能和大数据在地质工程细分领域的实践应用方向, 可以为相关人员野外地质工作提供有效助力。为了更好地促进岩性识别在专业领域的应用, 通过对巢湖北部山区的岩石图像采集、数据预处理、迁移学习、网络搭建、网络训练及模型测试等步骤, 实现了基于岩石图像的大数据深度学习识别; 并在归纳总结前人工作的基础上, 提出了多尺度岩性识别方法。根据岩石细观图像建立多尺度模型并赋予一定权重, 与岩石识别模型共同识别得到综合结果, 对岩石岩性整体识别的同时兼顾局部纹理、粒径等细观信息。研究结果表明, 本模型对岩石识别的适用性强, 多尺度方法对于提高识别结果的正确率具有一定的帮助, 模型的测试正确率达到95%以上, 能很好地识别岩石岩性。
- 深度学习 /
- 岩性识别 /
- 多尺度 /
- 权重 /
- 图像
Abstract:
Lithology recognition through artificial intelligence and big data can provide effective assistance to relevant personnel in field investigations.To better promote the application of lithology recognition in professional fields, the deep learning recognition of big data based on rock images were performed through the steps of rock image acquisition, data preprocessing, migration learning, network building, network training and model testing in the northern mountain area of Chaohu.Based on previous work, a multi-scale lithologic identification method is proposed. A multi-scale model is established and given a certain weight according to the rock meso image.The comprehensive results are obtained by identification with the rock identification model. The local texture, particle size and other mesoscale information were used in the overall identification of rock lithology.The results show that the multi-scale method is helpful to improve the identification results. The accuracy of the model is above 95%, which can well identify the rock lithology.
- deep learning /
- lithology recognition /
- multi-scale /
- weight /
- image

HTML全文

图 1 部分可视化训练样本

Figure 1. Part of visual training samples

下载: 全尺寸图片幻灯片

图 2 网络模型示意图(据文献[7]修改)

Figure 2. Schematic diagram of the network model

下载: 全尺寸图片幻灯片

图 3 模型网络架构图(据文献[17]修改)

Figure 3. Model network architecture

下载: 全尺寸图片幻灯片

图 4 模型网络训练过程曲线

Figure 4. Model network training process curve

下载: 全尺寸图片幻灯片

图 5 灰岩多尺度图片示意图

Figure 5. Multi-scale picture of limestone

下载: 全尺寸图片幻灯片

图 6 数据处理流程

Figure 6. Data processing process

下载: 全尺寸图片幻灯片

图 7 链条状灰岩

Figure 7. Chain limestone

下载: 全尺寸图片幻灯片

表 1 岩性数据集与数量

Table 1. Lithologic data set and quantity

类别名称	训练样本数/个	测试样本数/个
底部砾岩	385	96
砂岩	298	74
页岩	334	83
泥岩	298	74
链条状灰岩	32	8
其他灰岩	77	19
硅质岩	293	73
杂色页岩	354	89
方解石脉	187	47
碎石土	137	34

下载: 导出CSV

表 2 不同模型调试结果

Table 2. Debugging results of different models

迁移模型	训练集正确率/%	测试集正确率/%
ResNet152	95.54	87.50
ResNet101	98.21	84.82
ResNet50	97.32	91.07
InceptionV3	91.67	86.61
Xception	97.02	91.07

下载: 导出CSV

表 3 ResNet50模型调试结果

Table 3. Resnet50 model debugging results

批量大小	学习率
	0.01		0.001		0.000 5		0.000 1
	训练集正确率/%	测试集正确率/%	训练集正确率/%	测试集正确率/%	训练集正确率/%	测试集正确率/%	训练集正确率/%	测试集正确率/%
3	14.31	14.88	91.07	91.07	98.21	80.36	84.89	85.71
4	41.55	42.43	97.32	91.07	98.01	80.36	86.68	77.38
5	65.50	53.67	95.86	96.43	93.75	86.61	80.95	81.25
6	42.94	42.26	90.45	95.31	97.61	86.31	95.26	82.14
7	43.74	39.88	95.49	95.63	97.81	89.88	92.64	88.69

下载: 导出CSV

表 4 多尺度权重确定和综合识别概率

Table 4. Multi-scale weight determination and comprehensive recognition probability

权重	单尺度识别概率/%	多尺度识别概率/%	综合识别概率/%
0.05	95.26	97.19	95.65
0.10	94.23	97.86	94.98
0.15	96.12	98.28	96.55
0.20	97.89	99.13	98.14
0.25	94.89	98.02	95.52
0.30	95.30	97.59	95.76

下载: 导出CSV

参考文献(25)

[1]	李宁, 徐彬森, 武宏亮, 等. 人工智能在测井地层评价中的应用现状及前景[J]. 石油学报, 2021, 42(4): 508-522. https://www.cnki.com.cn/Article/CJFDTOTAL-SYXB202104008.htm Li N, Xu B S, Wu H L, et al. Application status and prospect of artificial intelligence in logging formation evaluation[J]. Acta Petrologica Sinica, 2021, 42(4): 508-522(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-SYXB202104008.htm
[2]	高婷, 杨洋, 何江. 人工智能在岩石图像上的应用综述[J]. 中国石油和化工标准与质量, 2020, 40(20): 98-100. doi: 10.3969/j.issn.1673-4076.2020.20.035 Gao T, Yang Y, He J. Application of artificial intelligence in rock image[J]. China Petroleum and Chemical Industry Standards and Quality, 2020, 40(20): 98-100(in Chinese with English abstract). doi: 10.3969/j.issn.1673-4076.2020.20.035
[3]	徐晗, 程丹仪, 徐永华, 等. 基于非开挖泥浆性能检测系统与弱监督学习的地层岩性识别[J]. 地质科技通报, 2021, 40(6): 293-301. doi: 10.19509/j.cnki.dzkq.2021.0629 Xu H, Cheng D Y, Xu Y H, et al. Formation lithology recognition based on trenchless mud performance detection system and weak supervision learning[J]. Bulletin of Geological Science and Technology, 2021, 40(6): 293-301(in Chinese with English abstract). doi: 10.19509/j.cnki.dzkq.2021.0629
[4]	张刚. 基于深度学习的遥感图像语义分割关键技术研究[D]. 成都: 中国科学院大学(中国科学院光电技术研究所), 2020. Zhang G. Research on key technologies of remote sensing image semantic segmentation based on deep learning[D]. Chengdu: University of Chinese Academy of Sciences(Institute of Optoelectronic Technology, Chinese Academy of Sciences), 2020(in Chinese with English abstract).
[5]	刘延保, 曹树刚, 刘玉成. 基于LS-SVM的岩石细观图像分析方法探讨[J]. 岩石力学与工程学报, 2008, 27(5): 1059-1065. doi: 10.3321/j.issn:1000-6915.2008.05.023 Liu Y B, Cao S G, Liu Y C. Discussion on analytical method for LS-SVM based mesoscopic rock images[J]. Chinese Journal of Rock Mechanics and Engineering, 2008, 27(5): 1059-1065(in Chinese with English abstract). doi: 10.3321/j.issn:1000-6915.2008.05.023
[6]	杨艳梅, 柳娜, 程国建, 等. 基于Spark平台的岩石图像聚类分析[J]. 西安石油大学学报: 自然科学版, 2016, 31(6): 114-118. doi: 10.3969/j.issn.1673-064X.2016.06.018 Yang Y M, Liu N, Cheng G J, et al. Clustering analysis of rock images based on spark platform[J]. Journal of Xi'an Shiyou University: Natural Science Edition, 2016, 31(6): 114-118(in Chinese with English abstract). doi: 10.3969/j.issn.1673-064X.2016.06.018
[7]	胡启成, 叶为民, 王琼, 等. 基于地质图像大数据的岩性识别研究[J]. 工程地质学报, 2020, 28(6): 1433-1440. https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202006030.htm Hu Q C, Ye W M, Wang Q, et al. Recognition of lithology with big data of geological images[J]. Journal of Engineering Geology, 2020, 28(6): 1433-1440(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202006030.htm
[8]	Patel A K, Chatterjee S. Computer vision-based limestone rock-type classification using probabilistic neural network[J]. Geoscience Frontiers, 2016, 7(1): 53-60. doi: 10.1016/j.gsf.2014.10.005
[9]	程国建, 刘丽婷. 深度学习算法应用于岩石图像处理的可行性研究[J]. 软件导刊, 2016, 15(9): 163-166. https://www.cnki.com.cn/Article/CJFDTOTAL-RJDK201609058.htm Cheng G J, Liu L T. Feasibility study of deep learning algorithm applied to rock image processing[J]. Software Guide, 2016, 15(9): 163-166(in Chinese). https://www.cnki.com.cn/Article/CJFDTOTAL-RJDK201609058.htm
[10]	魏珺洁. 深度学习在岩相识别中的应用[D]. 西安: 西安石油大学, 2020. Wei J J. Application of deep learning in lithofacies recognition[D]. Xi'an: Xi'an Shiyou University, 2020(in Chinese with English abstract).
[11]	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(2): 333-342. Zhang Y, Li M C, Han S. Automatic identification and classification in lithology based on deep learning in rock images[J]. Acta Petrologica Sinica, 2018, 34(2): 333-342(in Chinese with English abstract).
[12]	白林, 姚钰, 李双涛, 等. 基于深度学习特征提取的岩石图像矿物成分分析[J]. 中国矿业, 2018, 27(7): 178-182. https://www.cnki.com.cn/Article/CJFDTOTAL-ZGKA201807038.htm Bai L, Yao Y, Li S T, et al. Mineral composition analysis of rock image based on deep learning feature extraction[J]. China Mining Magazine, 2018, 27(7): 178-182(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-ZGKA201807038.htm
[13]	冯雅兴, 龚希, 徐永洋, 等. 基于岩石新鲜面图像与孪生卷积神经网络的岩性识别方法研究[J]. 地理与地理信息科学, 2019, 35(5): 89-94. https://www.cnki.com.cn/Article/CJFDTOTAL-DLGT201905016.htm Feng Y X, Gong X, Xu Y Y, et al. Lithology recognition based on fresh rock images and twins convolution neural network[J]. Geography and Geo-Information Science, 2019, 35(5): 89-94(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DLGT201905016.htm
[14]	吴磊, 吕国强, 薛治天, 等. 基于多尺度递归网络的图像超分辨率重建[J]. 光学学报, 2019, 39(6): 82-89. https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201906011.htm Wu L, Lü G Q, Xue Z T, et al. Super-resolution reconstruction of images based on multi-Scale recursive network[J]. Acta Optica Sinica, 2019, 39(6): 82-89(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-GXXB201906011.htm
[15]	王心源, 张广胜, 贾冠忠, 等. 安徽省巢湖野外地质实习基地建设与改革研究[J]. 中国地质教育, 2007(2): 47-51. https://www.cnki.com.cn/Article/CJFDTOTAL-JYDZ200702017.htm Wang X Y, Zhang G S, Jia G Z, et al. On the construction and reform of geological practice-teaching base in Chaohu, Anhui[J]. China Geological Education, 2007(2): 47-51(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-JYDZ200702017.htm
[16]	孟丹. 基于深度学习的图像分类方法研究[D]. 上海: 华东师范大学, 2017. Meng D. Research on image classification methods based on deep learning[D]. Shanghai: East China Normal University, 2017(in Chinese with English abstract).
[17]	李甫宪. 基于卷积神经网络的序列图像分类方法与研究[D]. 杭州: 浙江工业大学, 2020. Li F X. Research of sequence image classification methods based on convolutional neural network[D]. Hangzhou: Zhejiang University of Technology, 2020(in Chinese with English abstract).
[18]	Li G, Qiao Y, Zheng Y, et al. Semi-supervised learning based on generative adversarial network and its applied to lithology recognition[J]. IEEE Access, 2019, 7: 67428-67437.
[19]	Qi T T, Zhao H, Fu J. Research on the reservoir characteristics and lithology recognition based on asphalt sandstone[J]. IOP Conference Series: Earth and Environmental Science, 2019, 237(3): 032044.
[20]	李玮, 吴亮, 陈冠宇. 基于遥感分类的深度信念网络模型研究[J]. 地质科技情报, 2018, 37(2): 208-214. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201802028.htm Li W, Wu L, Chen G Y. Model of deep belief network based on remote sensing classification[J]. Geological Science and Technology Information, 2018, 37(2): 208-214(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ201802028.htm
[21]	沈焕锋, 李平湘, 张良培, 等. 图像超分辨率重建技术与方法综述[J]. 光学技术, 2009, 35(2): 194-203. https://www.cnki.com.cn/Article/CJFDTOTAL-GXJS200902008.htm Shen H F, Li P X, Zhang L P, et al. Overview on super resolution image reconstruction[J]. Optical Technique, 2009, 35(2): 194-203(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-GXJS200902008.htm
[22]	王建国, 刘国治, 周金山. 微波孔缝线性耦合函数研究[J]. 强激光与粒子束, 2003, 15(11): 1093-1099. https://www.cnki.com.cn/Article/CJFDTOTAL-QJGY200311013.htm Wang J G, Liu G Z, Zhou J S. Investigations on function for linear coupling of microwaves into slots[J]. High Power Laser & Particle Beams, 2003, 15(11): 1093-1099(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-QJGY200311013.htm
[23]	沈若兰, 黄英来, 温馨, 等. 基于Xception与ResNet50模型的蘑菇分类方法[J]. 黑河学院学报, 2020, 11(7): 181-184. https://www.cnki.com.cn/Article/CJFDTOTAL-HHXY202007062.htm Shen R L, Huang Y L, Wen X, et al. Mushroom classification based on Xception and ResNet50 models[J]. Journal of Heihe University, 2020, 11(7): 181-184(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-HHXY202007062.htm
[24]	张驰, 印桂生, 刘杰, 等. 一种基于卷积神经网络的多重风格人脸特征点检测方法: CN111027433A[P]. 2020. Zhang C, Yin G S, Liu J, et al. A multi style facial feature point detection method based on convolutional neural network: CN111027433A[P]. 2020(in Chinese).
[25]	徐晗, 姚孔轩, 程丹仪, 等. 基于非开挖随钻检测系统与随机森林的地层岩性识别[J]. 地质科技通报, 2021, 40(5): 272-280. doi: 10.19509/j.cnki.dzkq.2021.0039 Xu H, Yao K X, Cheng D Y, et al. Stratigraphic lithology identification based on trenchless drilling detection system and random forest[J]. Bulletin of Geological Science and Technology, 2021, 40(5): 272-280(in Chinese with English abstract). doi: 10.19509/j.cnki.dzkq.2021.0039