摘要
针对水深的卫星遥感反演模型多在底质均一假设下进行,较少考虑底质类型的问题,提出了一种底质分类视角下的反演模型。以南安达曼群岛、波照间岛和久米岛作为研究区域,结合实验室测量的典型底质反射率光谱,在Landsat 8 OLI上使用多种分类方法提取研究区域底质类型信息,发现支持向量机法得到的底质分类精度最佳。基于底质分类结果,利用多种模型构建不同底质类型的水深反演模型,并进一步对比未区分底质类型的水深反演模型。研究结果表明,基于底质分类的多元线性回归模型效果最佳,平均绝对误差为1.03 m,均方根误差为1.39 m,证明了区分底质类型建模可以提升水深反演精度。
In response to the problem that satellite remote sensing inversion models of bathymetry are mostly based on the assumption of homogeneous substrates with less consideration of substrate types,an inversion model from the perspective of substrate classification is proposed.In this paper,a variety of classification methods are used on Landsat 8 OLI to extract the substrate type information from the study area by combining the typical substrate reflectance spectra measured in the laboratory with the South Andaman Islands,Hateruma Island and Kume Island as the study area,and the best substrate classification accuracy is obtained by the support vector machine method.Based on the substrate classification results,bathymetric inversion models for different substrate types are constructed using various model forms and further compared with the bathymetric inversion models without differentiating the substrate types.The results show that the multiple linear regression model based on substrate classification is the best,with the mean absolute error 1.03 m and the root mean square error 1.39 m.This demonstrates that modelling with different substrate types can improve the accuracy of bathymetry inversion.
作者
刘瑾璐
孙德勇
焦红波
环宇
王胜强
LIU Jinlu;SUN Deyong;JIAO Hongbo;HUAN Yu;WANG Shengqiang(School of Marine Science,Nanjing University of Information Engineering,Nanjing 210044,China;Key Laboratory of Marine Dynamic Remote Sensing and Acoustics of Jiangsu Province,Nanjing 210044,China;Jiangsu Provincial Marine Environment Monitoring Engineering Technology Research Center,Nanjing 210044,China;National Marine Information Center,Tianjin 300017,China)
出处
《遥感信息》
CSCD
北大核心
2022年第5期101-107,共7页
Remote Sensing Information
基金
国家自然科学基金项目(42176179、41876203、42176181)
江苏省基础研究计划(自然科学基金)项目(BK20211289、BK20210667)。
关键词
Landsat
8
OLI
底质分类
经验方法
多元线性回归
水深反演
Landsat 8 OLI
sea bottom classification
empirical method
multiple linear regression
bathymetric inversion
