首页 | 期刊介绍 | 编委会 | 道德声明 | 投稿指南 | 常用下载 | 过刊浏览 | In English
引用本文:
【打印本页】   【HTML】   【下载PDF全文】   查看/发表评论  下载PDF阅读器  关闭
←前一篇|后一篇→ 过刊浏览    高级检索
本文已被:浏览 26次   下载 0  
分享到: 微信 更多
XGboost算法在多光谱遥感浅海水深反演中的应用
胡鹏, 赵露露, 高磊, 朱金山
山东科技大学测绘科学与工程学院
摘要:
在多光谱遥感浅海水深反演过程中,考虑到水体和底质影响,水深值和海水表面辐射亮度之间的线性关系不成立。本文以甘泉岛南部0~25m范围的沙质区域为研究区域,利用GeoEye-1多光谱遥感影像和多波束实测水深数据构建XGBoost非线性水深反演模型,研究了XGBoost算法用于水深反演的性能。以相关系数(R^2),均方误差(MSE)和平均绝对误差(MAE)作为评价指标,并与3种传统线性回归模型进行精度了对比分析。结果表明,XGBoost非线性水深反演模型拟合程度最好,精度优于线性回归模型,的R^2、MSE和MAE分别为0.99080.991、0.31250.33m和0.41680.44m,拟合程度最好,精度优于线性回归模型。为进一步探究各模型在不同水深的水深反演性能精度,将水深范围分成3段(0~8,8~15,15~25m)分别进行精度验证和误差分析。结果表明,XGBoost模型在任意各分段的精度均优于线性回归模型,MSE依次为0.56,0.14和0.43m反演水深的精度均优于线性回归模型,各分段的MSE依次为0.3304,0.1651和0.3682。
关键词:  光学浅海水深反演  XGBoost算法  非线性回归模型  底质类型
DOI:10.11759/hykx20191226002
分类号:
基金项目:地理信息工程国家重点实验室开放研究基金资助项目(SKLGIE2017-Z-3-3); 国家重点研发计划课题(协作)-极区海域水声环境观测与声场特性研究(2018YFC1405903); 测绘地理信息公益性行业科研专项(201512034)
Apply the XGBoost algorithm on multispectral shallow water bathymetry reteriving
HU Peng, ZHAO Lu-lu, GAO Lei, ZHU Jin-shan
Geomatics College,Shandong University of Science and Technology
Abstract:
For optical shallow water bathymetry retrieving, because of the influence of water type and sediment type, the relationship between the water depth and the sea surface reflectance is nonlinear. In this paper, we built a nonlinear depth inversion model using the XGBoost algorithm. The research area is the 0~25m sandy area around Ganquan island of South China Sea. GeoEye-1 multispectral data and in-situ multi beam data were used for this research to investigate the performance of XGBoost algorithm in depth inversion. The correlation coefficient (R^2), mean square error (MSE) and mean absolute error (MAE) were used to evaluation retrieved bathymetry results. The XGBoost bathymetry was compared with three linear regression models. The results show that the XGBoost nonlinear depth inversion model has the best fitting performance and better precision than the linear regression models, with R^2, MSE and MAE of 0.99080.991,0.31250.33m and 0.41680.44m. In order to further explore the performance of each model in different depths, the water depth is divided into 3 ranges (0~8, 8~15, 15~25m). Results show that, in each depth range the accuracy of XGBoost model is better than the linear regression models. The MSE in each depth range is 0.33040.56, 0.16510.14 and 0.36820.43m, respectively.
Key words:  optical shallow water depth inversion  XGBoost algorithm  nonlinear regression model  sediment type
版权所有 《海洋科学》编辑部 Copyright©2008 All Rights Reserved
主管单位:中国科学院 主办单位:中国科学院海洋研究所
地址:青岛市南海路七号  邮编:266071  电话:0532-82898755  E-mail:marinesciences@qdio.ac.cn
技术支持:北京勤云科技发展有限公司