Automatic classification of remote sensing lithology in the Huangshan Area of the Eastern Tianshan Mountains in Xinjiang Based on machine learning and analysis of its recognition accuracy
-
摘要:
研究目的 遥感岩性制图对于基础地质研究和矿产勘查均具有重要意义，针对传统岩性解译方法在复杂基岩区效率低、主观性强的问题，以新疆东天山黄山地区为研究区，构建融合光谱-空间特征的自动化分类模型，提升ASTER数据在基岩出露区的岩性识别精度，为矿产资源勘查提供技术支撑。
研究方法 提出分水岭分割与正则化极限学习机协同框架：①通过分水岭算法提取空间边界特征，建立空间约束规则库；②采用主成分分析和L2正则化优化光谱特征空间，简化ELM隐层结构；③设计最大投票机制融合光谱分类与空间约束结果。并与支持向量机(SVM)、最大似然法、马氏距离法等4类传统算法对比验证模型性能。
研究结果 实验表明：①融合模型总体精度达92.13%（Kappa=0.91），较SVM等传统分类方法精度大幅提高；②空间特征使花岗岩等相似岩性的区分精度提升；③特征降维后模型参数明显减少，分类时间大幅缩短。
结论 该模型通过多特征融合有效突破单一光谱分类瓶颈，为基岩区提供高精度、高效率的岩性识别新方案，可适配WorldView-3等数据并推广至类似基岩出露区域。
Abstract:Objective Remote sensing lithology mapping is of great significance for both basic geological research and mineral exploration. Aiming at the problems of low efficiency and strong subjectivity of traditional lithology interpretation methods in complex bedrock areas, this study takes the Huangshan area of the Eastern Tianshan Mountains in Xinjiang as the research area, aiming to construct an automatic classification model integrating spectroscopic and spatial characteristics. Improve the lithology identification accuracy of ASTER data in bedrock exposure areas and provide technical support for mineral resource exploration.
Methods Propose a collaborative framework of watershed segmentation and regularized extreme learning machine: ①Extract spatial boundary features through the watershed algorithm and establish a spatial constraint rule base; ②Principal component analysis and L2 regularization are adopted to optimize the spectral feature space and simplify the hidden layer structure of ELM. ③Design the maximum voting mechanism to integrate spectral classification and spatial constraint results. And compare and verify the model performance with four traditional algorithms such as Support Vector Machine (SVM), maximum likelihood method, and Markov distance method.
Results The experiments show that: ①The overall accuracy of the fusion model reaches 92.13% (Kappa=0.91), which is significantly improved compared with traditional classification methods such as SVM; ② Spatial characteristics improve the discrimination accuracy of similar rock types such as granite; ③After feature dimension reduction, the model parameters were significantly reduced and the classification time was greatly shortened.
Conclusions This model effectively breaks through the bottleneck of single spectral classification through multi−feature fusion, providing a new lithology identification scheme with high precision and high efficiency for the bedrock area. It can be adapted to data such as WorldView−3 and extended to similar bedrock exposure areas.
-
-
图 1 东天山构造单元与主要矿床分布(据张连昌等, 2021修改)
Figure 1.
表 1 研究区岩性类别及样本
Table 1. Lithology categories and samples of the study area
代号 所属地层或岩浆期 岩性 训练样本(像元数) 验证样本(像元数) Qhp1 第四系全新统 洪积砂砾岩 32 14 Qp-hpl 第四系全新统—上新统 洪积砂砾石 42 18 N2p 新近系葡萄沟组 粉砂质泥岩 35 15 C1g 上石炭统梧桐窝子组 硅质层凝灰岩 56 24 C2w 下石炭统干墩组 凝灰质硅质板岩 72 31 δ42a 华力西中期 黑云母花岗岩 48 21 γ42b 华力西中期 闪长岩 60 27 ν-Σ42a 华力西中期 基性—超基性岩体 35 15 
下载: 导出CSV
表 2 不同机器学习算法分类结果对比
Table 2. Comparison of classification results based on different machine learning algorithms
算法 SVM ELM 最大似然法 马氏距离法 分水岭法+ELM 分水岭法+PCA-RELM 总体分类精度 72.2% 78.2% 65.2% 59.7% 86.2% 92.3% Kappa系数 0.712 0.775 0.647 0.641 0.843 0.906 分类时间/s 975 439 589 387 569 290
下载: 导出CSV
-
[1] Ayerdi B, Marques I, Grana M. 2015. Spatially regularized semisupervised Ensembles of extreme learning machines for hyperspectral image segmentation[J]. Neurocomputing, 149: 373−386. doi: 10.1016/j.neucom.2014.01.068
[2] Bieniek A, Moga A. 2000. An efficient watershed algorithm based on connected components[J]. Pattern Recognition, 33: 907−916. doi: 10.1016/S0031-3203(99)00154-5
[3] Bhargava S, Somkuwar A. 2016. Estimation of noise removal techniques in medical imaging data−A review[J]. Journal of Medical Imaging and Health Informatics, 6: 875−884. doi: 10.1166/jmihi.2016.1797
[4] Cai L, Cheng G J, Pan H X. 2010. The application of extreme learning machine in lithology identification[J]. Computer Engineering and Design, 31: 2010−2012(in Chinese with English abstract).
[5] Cao W C, Li M, Liu J, et al. 2018. Accuracy evaluation of land cover classification data at different time phases Based on median division images[J]. Surveying, Mapping and Spatial Geographic Information, 41(11): 55−57(in Chinese with English abstract).
[6] Garea A S, Ordóñez Á, Heras D B, et al. 2016. HypeRvieW: An open source desktop applicationforhyperspectral remote-sensing data processing[J]. International Journal of Remote Sensing, 37(23): 5533−5550.
[7] Heras D B, Argüello F, Quesadabarriuso P. 2014. Exploring ELM−based spatial−spectral classification of hyperspectral images[J]. International Journal of Remote Sensing, 35(2): 401−423. doi: 10.1080/01431161.2013.869633
[8] Huang G B, Wang D H, Lan Y. 2011. Extreme learning machines: A survey[J]. International Journal of Machine Learning and Cybernetics, 2: 107−122. doi: 10.1007/s13042-011-0019-y
[9] Huang G B, Zhu Q Y, Siew C K. 2006. Extreme learning machine: Theory and applications[J]. Neurocomputing, 70: 489−501. doi: 10.1016/j.neucom.2005.12.126
[10] Huang G B, Zhu Q Y, Siew C K, et al. 2004. Extreme learning machine: A new learning scheme of feedforward neural networks[C]//IEEE International Joint Conference on Neural Networks. IEEE: 985−990.
[11] Huang H, Wang L H, Shi G Y. 2020. Spatial regularization manifold discriminant analysis method for hyperspectral image classification[J]. Acta Optica Sinica, 40(2): 181−191 (in Chinese with English abstract).
[12] He J X, Ren X Y, Chen S B, et al. 2021. Automatic classification of rock spectral characteristics based on fusion learning model[J]. Spectroscopy and Spectral Analysis, 41(1): 141−144(in Chinese with English abstract).
[13] Khelil M, Boudraa M, Kechida A, et al. 2005. Classification of defects by the SVM method and the principal component analysis (PCA)[J]. Enformatika, 9: 226−231.
[14] Li C, Song F. 2020. Research on remote sensing image classification method based on supervised classification[J]. Western Prospecting Engineering, 32(12): 159−162(in Chinese with English abstract).
[15] Long F, Ren J T. 2023. Research on GF−2 remote sensing image classification based on multi−classifier integration[J]. Science and Technology Innovation, 3: 19−24(in Chinese with English abstract).
[16] Li X T, Li C, Rahaman M M, et al. 2022. A comprehensive review of computer−aided whole−slide image analysis: from datasets to feature extraction, segmentation, classification and detection approaches[J]. Artificial Intelligence Review, 55: 4809−4878. doi: 10.1007/s10462-021-10121-0
[17] Maxwell A E, Warner T A, Fang F. 2018. Implementation of machine−learning classification in remote sensing: An applied review[J]. International Journal of Remote Sensing, 39: 2784−2817. doi: 10.1080/01431161.2018.1433343
[18] Mu D D, Liu L. 2019. A comparative study of ELM and SVM in supervised classification of hyperspectral remote sensing images[J]. Remote Sensing Technology and Application, 34(1): 115−124(in Chinese with English abstract).
[19] Mei J C, Liu L, et al. 2023. Research on lithology classification of ASTER images based on spectral−spatial characteristics: A case study of the Nishan area in Baixia, Beishan, Gansu Province[J]. Geological Review, 69(4): 2023040021(in Chinese with English abstract).
[20] Osuna E, Freund R, Girosi F, et al. 1997. Training support vector machines: An application to face detection[C]//IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE: 130−136.
[21] Shi Yu, Wang Y W, Wang J B, et al. 2022. Diagenetic and metallogenic mechanism of ore−bearing supermagnesium−iron rocks in the east and west Huangshan copper−nickel sulfide deposits of the Eastern Tianshan Mountains: Constrained by plagioclase composition[J]. Earth Science, 47(9): 3244−3257(in Chinese with English abstract).
[22] Sun K, Lu T D. 2017. Comparison of supervised classification methods in remote sensing image classification processing[J]. Jiangxi Science, 35(3): 367−371(in Chinese with English abstract).
[23] Tarabalka Y, Benediktsson J A, Chanussot J. 2009. Spectral–spatial classification of hyperspectral imagery based on partitional clustering techniques[J]. IEEE Transactions on Geoscience and Remote Sensing, 47: 2973−2987. doi: 10.1109/TGRS.2009.2016214
[24] Takemura R. 2022. Logic and majority voting[J]. Journal of Philosophical Logic, 51(2): 347−382. doi: 10.1007/s10992-021-09631-7
[25] Tauler R, Goupy J. 2011. Chemometrics and intelligent laboratory systems foreword[J]. Chemometrics and Intelligent Laboratory Systems, 106: 151−151.
[26] Wei A. 2015. Classifier design based on SVM algorithm[J]. Electronic Science and Technology, 28(4): 23−26(in Chinese with English abstract).
[27] Xia M Z. 2009. Genesis and mineralization of magnesium−iron − super−magnesium−iron rocks in the Huangshan Rock belt of the Eastern Tianshan Mountains in Xinjiang[D]. PhD Thesis of Chang’an University(in Chinese with English abstract).
[28] Xie W Y. 2022. Research on lithology identification and automatic classification based on GF−1 and Landsat 8 data: A case study of Changji area in the Eastern Tianshan Mountains of Xinjiang[D]. Master Thesis of Chang’an University(in Chinese with English abstract).
[29] Xu H Q, Tang F. 2013. New generation landsat series satellites: New features of Landsat 8 remote sensing images and their ecological and environmental significance[J]. Acta Ecologica Sinica, 33: 3249−3257(in Chinese with English abstract). doi: 10.5846/stxb201305030912
[30] Yin J, Qi C, Chen Q, et al. 2021. Spatial−spectral network for hyperspectral image classification: A 3−D CNN and Bi−LSTM framework[J]. Remote Sensing, 13(12): 2353. doi: 10.3390/rs13122353
[31] Yi X F, Wu J, Yu Gaohang. 2015. SVM kernel polarization parameter selection method based on adaptive gradient algorithm[J]. Journal of Gannan Normal University, 36(6): 23−27(in Chinese with English abstract).
[32] Zhang L Z, Dong Z G, Chen B. 2021. Important metallogenic belts, metallogenic systems and metallogenic laws in the Eastern Tianshan Mountains[J]. Chinese Journal of Earth Sciences and Environment, 43: 12−35(in Chinese with English abstract).
[33] Zhang J, Yao S K. 2011. The application of data mining in malicious program detection[J]. Computer Knowledge and Technology, 7(35): 9048−9049(in Chinese with English abstract).
[34] Zhu M Y, Li B Q, Fu H Z, et al. 2020. Research on SVM lithology classification based on multi−source data collaboration: A case study of Jianggalesayi area[J]. Uranium Geology, 36(4): 288−292, 317(in Chinese with English abstract).
[35] Zhang W S, Wang Z G. 2022. Missing data prediction of deep extreme learning machine based on improved sparrow algorithm[J]. Electronic Measurement Technology, 45(15): 63−67(in Chinese with English abstract).
[36] 蔡磊, 程国建, 潘华贤. 2010. 极限学习机在岩性识别中的应用[J]. 计算机工程与设计, 31: 2010−2012.
[37] 曹伟超, 李明, 刘江, 等. 2018. 基于中分影像的不同时相地表覆盖分类数据精度评价[J]. 测绘与空间地理信息, 41(11): 55−57. doi: 10.3969/j.issn.1672-5867.2018.11.015
[38] 贺金鑫, 任小玉, 陈圣波, 等. 2021. 融合学习模型的岩石光谱特征自动分类[J]. 光谱学与光谱分析, 41(1): 141−144.
[39] 黄鸿, 王丽华, 石光耀. 2020. 面向高光谱影像分类的空间正则化流形鉴别分析方法[J]. 光学学报, 40(2): 181−191.
[40] 李超, 宋方. 2020. 基于监督分类的遥感影像分类方法研究[J]. 西部探矿工程, 32(12): 159−162. doi: 10.3969/j.issn.1004-5716.2020.12.048
[41] 龙飞, 任金铜. 2023. 基于多分类器集成的GF-2遥感影像分类研究[J]. 科学技术创新, 3: 19−24. doi: 10.3969/j.issn.1673-1328.2023.02.006
[42] 梅佳成, 刘磊, 等. 2023. 基于光谱−空间特征的ASTER影像岩性分类研究——以甘肃北山白峡尼山地区为例[J]. 地质论评, 69(4): 2023040021.
[43] 牟多铎, 刘磊. 2019. ELM与SVM在高光谱遥感图像监督分类中的比较研究[J]. 遥感技术与应用, 34(1): 115−124.
[44] 石煜, 王玉往, 王京彬, 等. 2022. 东天山黄山东和黄山西铜镍硫化物矿床含矿超镁铁岩的成岩-成矿作用机制: 来自斜长石成分的约束[J]. 地球科学, 47(9): 3244−3257.
[45] 孙坤, 鲁铁定. 2017. 监督分类方法在遥感影像分类处理中的比较[J]. 江西科学, 35(3): 367−371.
[46] 危傲. 2015. 基于SVM算法的分类器设计[J]. 电子科技, 28(4): 23−26.
[47] 夏明哲. 2009. 新疆东天山黄山岩带镁铁—超镁铁质岩石成因及成矿作用[D]. 长安大学博士学位论文.
[48] 谢文扬. 2022. 基于GF-1和Landsat 8数据的岩性识别及自动分类研究——以新疆东天山昌吉地区为例[D]. 长安大学硕士学位论文.
[49] 徐涵秋, 唐菲. 2013. 新一代Landsat系列卫星: Landsat 8遥感影像新增特征及其生态环境意义[J]. 生态学报, 33: 3249−3257.
[50] 徐奕奕, 刘智琦, 刘琦. 2011. 基于改进的分水岭算法图像分割方法研究[J]. 计算机仿真, 28(9): 272−274. doi: 10.3969/j.issn.1006-9348.2011.09.068
[51] 易序峰, 吴坚, 喻高航. 2015. 基于自适应梯度算法的SVM核极化参数选择方法[J]. 赣南师范学院学报, 36(6): 23−27.
[52] 张敬, 姚书科. 2011. 数据挖掘在恶意程序检测中的应用[J]. 电脑知识与技术, 7(35): 9048−9049. doi: 10.3969/j.issn.1009-3044.2011.35.009
[53] 张连昌, 董志国, 陈博. 2021. 东天山重要成矿区带、成矿系统与成矿规律[J]. 地球科学与环境学报, 43: 12−35.
[54] 张文帅, 王占刚. 2022. 基于改进麻雀算法优化深度极限学习机的缺失数据预测[J]. 电子测量技术, 45(15): 63−67.
[55] 朱明永, 李炳谦, 付翰泽, 等. 2020. 基于多源数据协同的SVM岩性分类研究——以江尕勒萨依地区为例[J]. 铀矿地质, 36(4): 288−292. doi: 10.3969/j.issn.1000-0658.2020.04.007
-
图(9)
表(2)
计量
- 文章访问数: 53
- PDF下载数: 18
- 施引文献: 0