基于Sentinel-1A数据的多种机器学习算法识别冰山的比较

doi:10.11947/j.AGCS.2020.20190174

摘要/Abstract

摘要： 冰山识别对于海洋环境监测和船只安全运行等具有重要的意义，是北极航道开通和北极开发过程中的重要内容。采用合成孔径雷达（SAR）影像进行冰山识别具有独特的优势，多种机器学习算法均可用于SAR影像的冰山识别中。为了最大限度地发挥机器学习算法的性能，有必要对不同机器学习算法及其搭配使用的特征与特征标准化方法进行评估，从而进行最优冰山识别方法的选择。因此，本文基于Sentinel-1A SAR影像，采用多种机器学习方法、多种特征组合及多种特征标准化方法进行冰山识别，并比较各流程方法的识别性能差异。采用的机器学习算法包括贝叶斯分类器（Bayes）、反向神经网络（BPNN）、线性判别分析（LDA）、随机森林（RF）以及支持向量机（SVM）；特征标准化方法包括Min-max标准化、Z-score标准化及log函数标准化；数据集是含有12个SAR影像特征的969个冰山与非冰山样本，样本主要位于格陵兰岛东海岸。分类效果采用接收者操作特性（ROC）曲线下的面积（AUC）进行衡量。结果显示，最佳搭配下的RF的AUC值最高，达到了0.945，比最差的Bayes高出0.09。从识别率上来看，RF在冰山查全率为80%的情况下非冰山查全率达到92.6%，效果最好，比第2位的BPNN高出1.4%，比最差的Bayes高出2.6%；BPNN在冰山查全率为90%的情况下非冰山查全率达到87.4%，比第2位的RF高出0.8%，比最差的Bayes高出2.7%。上述结果表明，对冰山识别而言，选择最优的机器学习算法和最佳的特征与特征标准化方法都是十分重要的。

关键词: 冰山, 机器学习, Sentinel-1A, SAR

Abstract: Iceberg detection is of great significance for marine environmental monitoring and safe sailing of vessels. It is an important part of the construction of the Arctic channel and the exploitation of the Arctic. Iceberg detection using synthetic aperture radar (SAR) images has unique advantages. Many machine learning algorithms can be used in the recognition of icebergs in SAR images. In order to maximize the performance of machine learning algorithms, it is necessary to evaluate different machine learning algorithms and their matching feature and feature standardization methods, so as to select the optimal iceberg detection process method. Therefore, based on Sentinel-1A SAR image, this paper uses a variety of machine learning methods, a variety of feature combinations and a variety of feature standardization methods for iceberg detection, and compares the performance differences of each process method. Machine learning algorithms include Bayes classifier (Bayes), back propagation neural network (BPNN), linear discriminant analysis (LDA), random forest (RF) and support vector machine (SVM); feature standardization methods include Min-max standardization, Z-score standardization and log function standardization; data sets are comprised of 969 iceberg and non-iceberg samples with 12 SAR image features, located mainly on the east coast of Greenland. The classification result is measured by the area under the receiver operating characteristic (ROC) curve (AUC). The results show that the AUC value of RF with the best configuration is the highest, reaching 0.945, which is 0.09 higher than worst Bayes. In terms of detection rate, under the case of 80% iceberg recall rate, the non-iceberg recall rate of RF is 92.6%, which is the best, 1.4% higher than the second BPNN, 2.6% higher than the worst Bayes; under the case of 90% iceberg recall rate, the non-iceberg recall rate of BPNN is 87.4%, 0.8% higher than the second RF and 2.7% higher than the worst Bayes. The above results show that it is very important to select the best machine learning algorithm, the best features and feature standardization method for iceberg detection.

Key words: iceberg, machine learning, Sentinel-1A, SAR

中图分类号:

P237

肖湘文, 沈校熠, 柯长青, 周兴华. 基于Sentinel-1A数据的多种机器学习算法识别冰山的比较[J]. 测绘学报, 2020, 49(4): 509-521.

XIAO Xiangwen, SHEN Xiaoyi, KE Changqing, ZHOU Xinghua. Comparison of machine learning algorithms based on Sentinel-1A data to detect icebergs[J]. Acta Geodaetica et Cartographica Sinica, 2020, 49(4): 509-521.

参考文献

[1] BAMBER J, VAN DEN BROEKE M, ETTEMA J, et al. Recent large increases in freshwater fluxes from Greenland into the North Atlantic[J]. Geophysical Research Letters, 2012, 39(19):L19501.
[2] ENDERLIN E M, HOWAT I M, JEONG S, et al. An improved mass budget for the Greenland ice sheet[J]. Geophysical Research Letters, 2014, 41(3):866-872.
[3] RIGNOT E, VELICOGNA I, VAN DEN BROEKE M R, et al. Acceleration of the contribution of the Greenland and Antarctic ice sheets to sea level rise[J]. Geophysical Research Letters, 2011, 38(5):L05503.
[4] DUPRAT L P A M, BIGG G R, WILTON D J. Enhanced Southern Ocean marine productivity due to fertilization by giant icebergs[J]. Nature Geoscience, 2016, 9(3):219-221.
[5] MARTIN T, ADCROFT A. Parameterizing the fresh-water flux from land ice to ocean with interactive icebergs in a coupled climate model[J]. Ocean Modelling, 2010, 34(3-4):111-124.
[6] SILVA T A M, BIGG G R. Computer-based identification and tracking of Antarctic icebergs in SAR images[J]. Remote Sensing of Environment, 2005, 94(3):287-297.
[7] WESCHE C, DIERKING W. Iceberg signatures and detection in SAR images in two test regions of the Weddell Sea, Antarctica[J]. Journal of Glaciology, 2012, 58(208):325-339.
[8] LANE K, POWER D, CHAKRABORTY I, et al. RADARSAT-1 synthetic aperture radar iceberg detection performance ADRO-2 A223[C]//Proceedings of IEEE International Geoscience and Remote Sensing Symposium. Toronto:IEEE, 2002.
[9] LANE K, POWER D, YOUDEN J, et al. Validation of synthetic aperture radar for iceberg detection in sea ice[C]//Proceedings of 2004 IEEE International Geoscience and Remote Sensing Symposium. Anchorage:IEEE, 2004.
[10] HOWELL C, MILLS J, POWER D, et al. A multivariate approach to iceberg and ship classification in HH/HV ASAR data[C]//Proceedings of 2006 IEEE International Symposium on Geoscience and Remote Sensing. Denver:IEEE, 2006.
[11] WESCHE C, DIERKING W. Iceberg signatures and detection in SAR images in two test regions of the Weddell Sea, Antarctica[J]. Journal of Glaciology, 2012, 58(208):325-339.
[12] CLEVE C, KELLY M, KEARNS F R, et al. Classification of the wildland-urban interface:a comparison of pixel- and object-based classifications using high-resolution aerial photography[J]. Computers, Environment and Urban Systems, 2008, 32(4):317-326.
[13] WHITESIDE T G, BOGGS G S, MAIER S W. Comparing object-based and pixel-based classifications for mapping savannas[J]. International Journal of Applied Earth Observation and Geoinformation, 2011, 13(6):884-893.
[14] FISCELLA B, GIANCASPRO A, NIRCHIO F, et al. Oil spill detection using marine SAR images[J]. International Journal of Remote Sensing, 2000, 21(18):3561-3566.
[15] STATHAKIS D, TOPOUZELIS K, KARATHANASSI V. Large-scale feature selection using evolved neural networks[C]//Proceedings of the SPIE 6365, Image and Signal Processing for Remote Sensing Ⅻ. Stockholm:SPIE, 2006:636513.
[16] BRENNING A, KADEN K, ITZEROTT S. Comparing classifiers for crop identification based on multitemporal Landsat TM/ETM data[C]//Proceedings of the 2nd Workshop of the EARSeL SIG on Land Use and Land Cover. Bonn:[s.n.], 2006:30.
[17] BRENNING A. Benchmarking classifiers to optimally integrate terrain analysis and multispectral remote sensing in automatic rock glacier detection[J]. Remote Sensing of Environment, 2009, 113(1):239-247.
[18] 赵诣, 蒋弥. 极化SAR参数优化与光学波谱相结合的面向对象土地覆盖分类[J]. 测绘学报, 2019, 48(5):609-617. DOI:10.11947/j.AGCS.2019.20170746. ZHAO Yi, JIANG Mi. Integration of SAR polarimetric parameters and multi-spectral data for object-based land cover classification[J]. Acta Geodaetica et Cartographica Sinica, 2019, 48(5):609-617. DOI:10.11947/j.AGCS.2019.20170746.
[19] 周志华. 机器学习[M]. 北京:清华大学出版社, 2016:23-47. ZHOU Zhihua. Machine learning[M]. Beijing:Tsinghua University Press, 2016:23-47.
[20] WILLIS C J, MACKLIN J T, PARTINGTON K C, et al. Iceberg detection using ERS-1 synthetic aperture radar[J]. International Journal of Remote Sensing, 1996, 17(9):1777-1795.
[21] 赵泉华, 郭世波, 李晓丽, 等. 利用目标分解特征的全极化SAR海冰分类[J]. 测绘学报, 2018, 47(12):1609-1620. DOI:10.11947/j.AGCS.2018.20170551. ZHAO Quanhua, GUO Shibo, LI Xiaoli, et al. Polarimetric SAR sea ice classification based on target decompositional features[J]. Acta Geodaetica et Cartographica Sinica, 2018, 47(12):1609-1620. DOI:10.11947/j.AGCS.2018.20170551.
[22] BLASCHKE T, HAY G J, KELLY M, et al. Geographic object-based image analysis-towards a new paradigm[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2014, 87:180-191.
[23] 明冬萍, 邱玉芳, 周文. 遥感模式分类中的空间统计学应用——以面向对象的遥感影像农田提取为例[J]. 测绘学报, 2016, 45(7):825-833. DOI:10.11947/j.AGCS.2016.20150520. MING Dongping, QIU Yufang, ZHOU Wen. Applying spatial statistics into remote sensing pattern recognition:with case study of cropland extraction based on GeoBIA[J]. Acta Geodaetica et Cartographica Sinica, 2016, 45(7):825-833. DOI:10.11947/j.AGCS.2016.20150520.
[24] 张仙, 明冬萍. 面向地学应用的遥感影像分割评价[J]. 测绘学报, 2015, 44(S1):108-116. DOI:10.11947/j.AGCS.2015.F041. ZHANG Xian, MING Dongping. Geo-application oriented evaluations of remote sensing image segmentation[J]. Acta Geodaetica et Cartographica Sinica, 2015, 44(S1):108-116. DOI:10.11947/j.AGCS.2015.F041.
[25] 齐丽娜, 张博, 王战凯. 最大类间方差法在图像处理中的应用[J]. 无线电工程, 2006, 36(7):25-26, 44. QI Lina, ZHANG Bo, WANG Zhankai. Application of the OTSU method in image processing[J]. Radio Engineering of China, 2006, 36(7):25-26, 44.
[26] 郭臻, 陈远知. 图像阈值分割算法研究[J]. 中国传媒大学学报(自然科学版), 2008, 15(2):77-82. GUO Zhen, CHEN Yuanzhi. Research of thresholding methods for image segmentation[J]. Journal of Communication University of China Science and Technology, 2008, 15(2):77-82.
[27] XU Linlin, LI J, BRENNING A. A comparative study of different classification techniques for marine oil spill identification using RADARSAT-1 imagery[J]. Remote Sensing of Environment, 2014, 141:14-23.
[28] METZ C E. Basic principles of ROC analysis[J]. Seminars in Nuclear Medicine, 1978, 8(4):283-298.
[29] STROBL C, BOULESTEIX A L, ZEILEIS A, et al. Bias in random forest variable importance measures:illustrations, sources and a solution[J]. BMC Bioinformatics, 2007, 8(1):25.
[30] GONG Jianya. Photogrammetry and deep learning[J]. Journal of Geodesy and Geoinformation Science, 2018, 1(1):1-15.