Automatic Lithology Classification Based on Deep Features Using Dual Polarization SAR Images
-
摘要: 基于像元基元、极化合成孔径雷达（Synthetic Aperture Radar，SAR）数据和传统机器学习算法的岩性分类方法，易受SAR图像固有斑点噪声影响，精度不高.为了降低噪声的影响，本研究以大尺度像元邻域为基元，用于表征地表地质体的遥感图像特征和岩性语义信息；采用高分三号双极化SAR数据进行极化分解构建3通道假彩色合成影像；然后采用深度卷积神经网络（Deep Convolutional Neural Network，DCNN）迁移学习的方法，提取有效的深度特征表示，分别实现5 m和15 m两种空间分辨率下岩性遥感自动分类.结果表明：基于不同分辨率数据和不同DCNN算法，岩性遥感自动分类的总精度均大于80%，最高精度达到91%.基于大尺度像元邻域和DCNN迁移学习方法，能够实现基于SAR数据的高精度岩性分类.Abstract: The lithology classification method based on pixel primitives, polarimetric synthetic aperture radar (SAR) data and traditional machine learning algorithm is easy to be affected by the inherent speckle noise, and the accuracy is not high. In order to reduce the effect of image noise, the neighborhood of large-scale pixels is considered as the primitive to characterize the spatial aggregation characteristics of surface geological units and the corresponding lithologic semantic information. Using GaoFen-3 dual polarization data, the polarization decomposition is carried out first, and a 3-channel color composite image is constructed as the input data of the subsequent model. Then, the deep convolutional neural network (DCNN) based migration learning method is used to extract the effective deep feature representation, so as to realize the automatic lithology classification under 5 m and 15 m spatial resolution conditions. The experiment results show that based on different resolution data and different DCNN algorithms, the total accuracy of automatic lithology classification is greater than 80%, and the highest accuracy is 91%. Generally, based on large-scale pixel neighborhood and DCNN migration learning method, high-precision lithology classification based on SAR data can be realized. The lithology remote sensing dataset based on dual polarization SAR created in this paper can also be used as the benchmark of lithology classification based on artificial intelligence.
-
Key words:
- remote sensing /
- lithology classification /
- transfer learning /
- deep convolutional neural network /
- GaoFen-3 /
- SAR
-
图 2 研究区岩性和地表水分布图及野外验证点与照片
Fig. 2. Constructed lithology and surface water distribution, field verification points and photos in the study area
图 6 基于迁移学习和大尺度邻域的极化SAR岩性分类模型
Fig. 6. Polarimetric SAR lithology classification model based on migration learning and large-scale neighborhood
图 7 CUG_GF3SD_15真实标签图和分类结果
a.真值标签；b.VGG16；c.ResNet；d.DenseNet
Fig. 7. Classification results of CUG_GF3SD_15
图 8 CUG_GF3SD_15分辨率混淆矩阵
a.VGG16；b. DenseNet；c. ResNet.0.板岩；1.花岗闪长岩；2.黄土；3.片岩；4.水体；5.松散堆积物
Fig. 8. Confusion matrix of CUG_GF3SD_15
图 9 CUG_GF3SD_5分类结果
a.真值标签；b.VGG16；c.ResNet；d.DenseNet
Fig. 9. Classification results of CUG_GF3SD_5
图 10 CUG_GF3SD_5分辨率混淆矩阵
a.VGG16；b.DenseNet；c.ResNet. 0.板岩；1.花岗闪长岩；2.黄土；3.片岩；4.水体；5.松散堆积物
Fig. 10. Confusion matrix of CUG_GF3SD_5
图 11 CUG_GF3SD_M5分类结果
a.真值标签；b.VGG16；c.ResNet；d.DenseNet
Fig. 11. Classification results of CUG_GF3SD_M5
图 12 CUG_GF3SD_M5分辨率混淆矩阵
a.VGG16；b.DenseNet；c.ResNet. 0.板岩；1.花岗闪长岩；2.黄土；3.片岩；4.水体；5.松散堆积物
Fig. 12. Confusion matrix of CUG_GF3SD_M5
表 1 GF⁃3卫星有效载荷技术指标（刘杰和张庆君，2018）
Table 1. Main technical index of GF⁃3 satellite payload (Liu and Zhang, 2018)
成像模式名称 分辨率(m) 幅宽(km) 极化方式 滑块聚束（SL） 1 10 单极化 条带成像模式 UFS 3
5
10
25
8
2530 单极化 FSI 50 双极化 FSII 100 双极化 SS 130 双极化 QPSI 30 全极化 QPSII 40 全极化 扫描成像模式 NSC 50 300 双极化 WSC 100 500 双极化 GLO 500 650 双极化 波成像模式（WAV） 10 5 全极化 扩展入射角（EXT） 低入射角 25 130 双极化 高入射角 25 80 双极化 
下载: 导出CSV
表 2 不同数据集采样点数量在相应类别的比例
Table 2. Proportion of corresponding categories of sampling points of different data sets
数据集 松散堆积物(%) 板岩(%) 水体(%) 片岩(%) 花岗闪长岩(%) 黄土(%) CUG_GF3SD_5 1.85 4.48 51.51 16.35 0.91 34.96 CUG_GF3SD_15 1.85 4.48 51.51 16.35 0.91 34.96 CUG_GF3SD_M5 0.13 0.34 3.65 1.09 0.06 3.44
下载: 导出CSV
表 3 不同分辨率下各特征提取网络分类精度
Table 3. Classification accuracy of backbone networks under different resolutions
特征提取网络 数据集 松散堆积物 板岩 水体 片岩 花岗闪长岩 黄土 OA Kappa VGG16 CUG_GF3SD_15 0.79 0.86 0.87 0.65 0.94 0.69 0.89 0.78 CUG_GF3SD_5 0.86 0.65 0.93 0.89 0.87 0.82 0.86 0.71 DenseNet121 CUG_GF3SD_15 0.76 0.57 0.86 0.55 0.94 0.62 0.88 0.73 CUG_GF3SD_5 0.86 0.64 0.93 0.89 0.87 0.82 0.85 0.68 ResNet101v2 CUG_GF3SD_15 0.64 0.47 0.78 0.35 0.95 0.67 0.85 0.66 CUG_GF3SD_5 0.84 0.60 0.92 0.82 0.84 0.75 0.84 0.67
下载: 导出CSV
表 4 CUG_GF3SD_M5分辨率各特征提取网络分类精度
Table 4. CUG_GF3SD_M5 classification accuracy of backbone networks
特征提取网络（5 m） 松散堆积物 板岩 水体 片岩 花岗闪长岩 黄土 OA Kappa VGG16 0.90 0.84 0.93 0.89 0.91 0.85 0.91 0.80 DenseNet121 0.90 0.84 0.93 0.91 0.90 0.82 0.90 0.78 ResNet101v2 0.89 0.82 0.93 0.87 0.90 0.80 0.90 0.77
下载: 导出CSV
-
Chen, G.X., Li, P.P., Liu, S.D., et al., 2018. Extraction and 3D Visualization of Surface Lithology Based on GF-1 Satellite Images. Geography and Geo-Information Science, 34(5): 31-36, 2 (in Chinese with English abstract). doi: 10.3969/j.issn.1672-0504.2018.05.006 Cheng, S.Y., Chen, J., Lin, H.X., et al., 2021. Application of Geometric Precision Correction Based on High-Resolution Remote Sensing Image in 1: 50 000 Geological Mapping. Geological Bulletin of China, 40(4): 520-526 (in Chinese with English abstract). Dai, C., Li, W., Wang, D., et al., 2021. Active Landslide Detection Based on Sentinel-1 Data and InSAR Technology in Zhouqu County, Gansu Province, Northwest China. Journal of Earth Science, 32(5): 1092-1103. doi: 10.1007/s12583-020-1380-0 Deng, J., Dong, W., Socher, R., et al., 2009. Imagenet: A Large-Scale Hierarchical Image Database. 2009 IEEE Conference on Computer Vision and Pattern Recognition, Miami Beach, 248-255. Dong, X.F., Gan, F.P., Li, N., et al., 2020. Fine Mineral Identification of GF-5 Hyperspectral Image. Journal of Remote Sensing, 24(4): 454-464 (in Chinese with English abstract). Fu, G.M., Yan, J.Y., Zhang, K., et al., 2017. Current Status and Progress of Lithology Identification Technology. Progress in Geophysics, 32(1): 26-40 (in Chinese with English abstract). He, K., Zhang, X., Ren, S., et al., 2016. Identity Mappings in Deep Residual Networks. Lecture Notes in Computer Science, 630-645. Huang, G., Liu, Z., Laurens, V.D.M., et al., 2017. Densely Connected Convolutional Networks. 2017 IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, 4700-4708. Kalia, A. C., Spreckels, V., Lege, T., 2021. Comparison of L-and C-B and SAR Data in the Saar Mining District, Germany. EGU General Assembly Conference Abstracts, Online, EGU21-12736. https://doi.org/10.5194/egusphere-egu21-12736 Kim, J. W., Lu, Z., Kaufmann, J., 2019. Evolution of Sinkholes over Wink, Texas, Observed by High-Resolution Optical and SAR Imagery. Remote Sensing of Environment, 222: 119-132. doi: 10.1016/j.rse.2018.12.028 Krizhevsky, A., Sutskever, I., Hinton, G.E., 2012. Imagenet Classification with Deep Convolutional Neural Networks. Advances in Neural Information Processing Systems, 25: 1097-1105. Li, N., Dong, X.F., Gan, F.P., et al., 2021. Application of Hyperspectral Remote Sensing Technology to Regional Geological Survey and Mapping in Bedrock Area. Geological Bulletin of China, 40(1): 13-21 (in Chinese with English abstract). Li, P., Li, Z., Dai, K., et al., 2021. Reconstruction and Evaluation of DEMs from Bistatic TanDEM-X SAR in Mountainous and Coastal Areas of China. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 14: 5152-5170. doi: 10.1109/JSTARS.2021.3073782 Liu, F., Jiao, L., Tang, X., 2019. Task-Oriented GAN for PolSAR Image Classification and Clustering. IEEE Transactions on Neural Networks and Learning Systems, 30(9): 2707-2719. https://doi.org/10.1109/tnnls.2018.2885799 Liu, J., Zhang, Q.J., 2018. GF-3 Satellite and Its Application. Satellite Application, (6): 12-16 (in Chinese with English abstract). doi: 10.3969/j.issn.1674-9030.2018.06.006 Liu, L., Zhou, J., Jiang, D., et al., 2014. Lithological Discrimination of the Mafic-Ultramafic Complex, Huitongshan, Beishan, China: Using ASTER Data. Journal of Earth Science, 25(3): 529-536. https://doi.org/10.1007/s12583-014-0437-3 Liu, X., Zhao, C., Zhang, Q., et al., 2021. Integration of Sentinel-1 and ALOS/PALSAR-2 SAR Datasets for Mapping Active Landslides along the Jinsha River Corridor, China. Engineering Geology, 284: 106033. doi: 10.1016/j.enggeo.2021.106033 Lu, Y., Yang, C., Meng, Z., 2021. Lithology Discrimination Using Sentinel-1 Dual-Pol Data and SRTM Data. Remote Sensing, 13(7): 1280. doi: 10.3390/rs13071280 Pei, J.F., Huo, W.B., Wang, C.W., et al., 2021. Multiview Deep Feature Learning Network for SAR Automatic Target Recognition. Remote Sensing, 13(8): 1455. doi: 10.3390/rs13081455 Saepuloh, A., Bakker, E., Suminar, W., 2017. The Significance of SAR Remote Sensing in Volcano-Geology for Hazard and Resource Potential Mapping. AIP Conference Proceedings, 1857(1): 070005. https://doi.org/10.1063/1.4987093 Sekandari, M., Masoumi, I., Pour, A. B., et al., 2022. ASTER and WorldView-3 Satellite Data for Mapping Lithology and Alteration Minerals Associated with Pb-Zn Mineralization. Geocarto International, 37(6): 1782-1812. doi: 10.1080/10106049.2020.1790676 Simonyan, K., Zisserman, A., 2015. Very Deep Convolutional Networks for Large-Scale Image Recognition. International Conference on Learning Representations (ICLR), San Diego, 1-14. Wang, R.S., 2008. On the Development Strategy of Remote Sensing Technology in Geology. Remote Sensing for Land & Resources, 85(1): 1-12, 42 (in Chinese with English abstract). Wang, R.S., Xiong, S.Q., Nie, H.F., et al., 2011. Remote Sensing Technology and Its Application in Geological Exploration. Acta Geologica Sinica, 85(11): 1699-1743 (in Chinese with English abstract). Wang, W., Ren, X., Zhang, Y., et al., 2018. Deep Learning Based Lithology Classification Using Dual-Frequency Pol-SAR Data. Applied Sciences, 8(9): 1513. doi: 10.3390/app8091513 Wang, X.S., Chen, E.X., Li, Z.Y., et al., 2015. Multi-Temporal and Dual-Polarization Interferometric SAR for Land Cover Type Classification. Acta Geodaetica et Cartographica Sinica, 44(5): 533-540 (in Chinese with English abstract). Xia, W.H., Xu, H., 2021. Study on Geological Hazards in Mining Areas Based on D-InSAR Technology. Geomatics & Spatial Information Technology, 44(2): 125-129, 134 (in Chinese with English abstract). doi: 10.3969/j.issn.1672-5867.2021.02.031 Xie, M., Zhang, Q., Chen, S., et al., 2015. A Lithological Classification Method from Fully Polarimetric SAR Data Using Cloude-Pottier Decomposition and SVM. AOPC 2015: Optical and Optoelectronic Sensing and Imaging Technology. International Society for Optics and Photonics, 9674: 967405. Yang, T., Gong, H.L., Li, X.J., et al., 2010. Application of SAR to Remote Sensing of Geological Disasters. Journal of Natural Disasters, 19(5): 42-48 (in Chinese with English abstract). Yu, X.C., Zhou, X., Kang, Z.J., et al., 2012. Hierarchical Classification of Rock and Soil Based on Characteristic Multi-Band Image. Journal of Jilin University (Earth Science Edition), 42(6): 1825-1833 (in Chinese with English abstract). Yu, Y.F., Yang, J.Z., Chen, S.B., et al., 2015. Lithologic Classification from Remote Sensing Images Based on Spectral Index. Earth Science, 40(8): 1415-1419 (in Chinese with English abstract). Yuan, W., Yan, M., Liu, S., 2016. Application of Radar and Optical Remote Sensing Data in Lithologic Classification and Identification. 2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS), Beijing, 6370-6373. Zhang, B., 2018. Remotely Sensed Big Data Era and Intelligent Information Extraction. Geomatics and Information Science of Wuhan University, 43(12): 1861-1871 (in Chinese with English abstract). Zhang, C.F., Hao, L.N., Wang, S.J., et al., 2020. Geological Units Classification with Texture-Spectral Synergy of Multi-Sourced Remote Sensing Images. Earth Science, 45(5): 1844-1854 (in Chinese with English abstract). Zhang, W., Li, Y.Y., Zhang, T.L., et al., 2019. Remote Sensing Interpretation of Landslide Geological Hazards in High Vegetation Coverage Area Based on Hazard Sensitivity Analysis. Safety and Environmental Engineering, 26(3): 28-35 (in Chinese with English abstract). Zhang, W., Lin, J., Chen, L., et al., 2014. Geological Information Extraction Using Polarimetric SAR Based on Polarization Decomposition. Remote Sensing Information, 29(1): 10-14 (in Chinese with English abstract). doi: 10.3969/j.issn.1000-3177.2014.01.003 Zhang, Z., Wang, H., Feng. X., et al., 2017. Complex-Valued Convolutional Neural Network and Its Application in Polarimetric SAR Image Classification. IEEE Transactions on Geoscience and Remote Sensing, (12): 1-12. Zhang, Z.Y., Liu, C., 2020. Terrain Classification of Pol-SAR Based on Dilated Convolution and Polarization Decomposition. Microelectronics & Computer, 37(12): 70-76 (in Chinese). Zheng, S., Fu, B.H., 2013. Lithological Mapping of Granitiods in the Western Junggar from ASTER SWIR-TIR Multispectral Data: Case Study in Karamay Pluton, Xinjiang. Acta Petrologica Sinica, 29(8): 2936-2948 (in Chinese with English abstract). Zhou, Y., Wang, H., Xu., F., et al., 2017. Polarimetric SAR Image Classification Using Deep Convolutional Neural Networks. IEEE Geoscience & Remote Sensing Letters, 13(12): 1935-1939. 陈国旭, 李盼盼, 刘盛东, 等, 2018. 基于高分一号卫星遥感影像的地表岩性特征提取及三维可视化. 地理与地理信息科学, 34(5): 31-36, 2. doi: 10.3969/j.issn.1672-0504.2018.05.006 程三友, 陈静, 林海星, 等, 2021. 高分辨率遥感图像几何精校正在高山峡谷区1: 5万地质填图中的应用. 地质通报, 40(4): 520-526. https://www.cnki.com.cn/Article/CJFDTOTAL-ZQYD202104008.htm 董新丰, 甘甫平, 李娜, 等, 2020. 高分五号高光谱影像矿物精细识别. 遥感学报, 24(4): 454-464. 付光明, 严加永, 张昆, 等, 2017. 岩性识别技术现状与进展. 地球物理学进展, 32(1): 26-40. https://www.cnki.com.cn/Article/CJFDTOTAL-DQWJ201701004.htm 李娜, 董新丰, 甘甫平, 等, 2021. 高光谱遥感技术在基岩区区域地质调查填图中的应用. 地质通报, 40(1): 13-21. https://www.cnki.com.cn/Article/CJFDTOTAL-ZQYD202101003.htm 刘杰, 张庆君, 2018. 高分三号卫星及应用概况. 卫星应用, (6): 12-16. https://www.cnki.com.cn/Article/CJFDTOTAL-WXYG201806005.htm 王润生, 2008. 遥感地质技术发展的战略思考. 国土资源遥感, 85(1): 1-12, 42. https://www.cnki.com.cn/Article/CJFDTOTAL-GTYG200801000.htm 王润生, 熊盛青, 聂洪峰, 等, 2011. 遥感地质勘查技术与应用研究. 地质学报, 85(11): 1699-1743. https://www.cnki.com.cn/Article/CJFDTOTAL-DZXE201111003.htm 王馨爽, 陈尔学, 李增元, 等, 2015. 多时相双极化合成孔径雷达干涉测量土地覆盖分类方法. 测绘学报, 44(5): 533-540. https://www.cnki.com.cn/Article/CJFDTOTAL-CHXB201505011.htm 夏伟华, 徐涵, 2021. 基于D-InSAR技术的矿区地质灾害研究. 测绘与空间地理信息, 44(2): 125-129, 134. https://www.cnki.com.cn/Article/CJFDTOTAL-DBCH202102031.htm 杨涛, 宫辉力, 李小娟, 等, 2010. 成像雷达遥感地质灾害应用. 自然灾害学报, 19(5): 42-48. https://www.cnki.com.cn/Article/CJFDTOTAL-ZRZH201005007.htm 余先川, 周鑫, 康增基, 等, 2012. 一种基于多特征波段岩土层次分类方法. 吉林大学学报(地球科学版), 42(6): 1825-1833. https://www.cnki.com.cn/Article/CJFDTOTAL-CCDZ201206026.htm 于亚凤, 杨金中, 陈圣波, 等, 2015. 基于光谱指数的遥感影像岩性分类. 地球科学, 40(8): 1415-1419. doi: 10.3799/dqkx.2015.127 张兵, 2018. 遥感大数据时代与智能信息提取. 武汉大学学报(信息科学版), 43(12): 1861-1871. https://www.cnki.com.cn/Article/CJFDTOTAL-WHCH201812013.htm 张翠芬, 郝利娜, 王少军, 等, 2020. 多源遥感数据图谱协同岩石单元分类方法. 地球科学, 45(5): 1844-1854. doi: 10.3799/dqkx.2019.168 张为, 李远耀, 张泰丽, 等, 2019. 基于孕灾敏感性分析的高植被覆盖区滑坡地质灾害遥感解译. 安全与环境工程, 26(3): 28-35. https://www.cnki.com.cn/Article/CJFDTOTAL-KTAQ201903005.htm 张微, 林健, 陈玲, 等, 2014. 基于极化分解的极化SAR数据地质信息提取方法研究. 遥感信息, 29(1): 10-14. https://www.cnki.com.cn/Article/CJFDTOTAL-YGXX201401005.htm 张泽宇, 刘畅, 2020. 基于极化分解和膨胀卷积的极化SAR地物分类. 微电子学与计算机, 37(12): 70-76. https://www.cnki.com.cn/Article/CJFDTOTAL-WXYJ202012014.htm 郑硕, 付碧宏, 2013. 基于ASTER SWIR-TIR多光谱数据的西准噶尔花岗岩类岩性信息提取与识别——以克拉玛依岩体为例. 岩石学报, 29(8): 2936-2948. https://www.cnki.com.cn/Article/CJFDTOTAL-YSXB201308029.htm -
点击查看大图
图(12) / 表(4)
计量
- 文章访问数: 1415
- HTML全文浏览量: 549
- PDF下载量: 80
- 被引次数: 0
