测绘通报 ›› 2022, Vol. 0 ›› Issue (7): 33-37.doi: 10.13474/j.cnki.11-2246.2022.0199

• 学术研究 • 上一篇    下一篇

CatBoost模型在水深反演中的应用

孔瑞瑶1, 谢涛1,2, 马明3, 孔瑞林4   

  1. 1. 南京信息工程大学遥感与测绘工程学院, 江苏 南京 210044;
    2. 青岛海洋科学与技术国家实验室区域海洋动力学与数值模拟功能实验室, 山东 青岛 266237;
    3. 北京应用气象研究所, 北京 100029;
    4. 西北工业大学软件学院, 陕西 西安 710129
  • 收稿日期:2022-01-25 出版日期:2022-07-25 发布日期:2022-07-28
  • 通讯作者: 谢涛。E-mail:xietao@nuist.edu.cn
  • 作者简介:孔瑞瑶(1997—),女,硕士生,主要研究方向为海洋遥感。E-mail:kongruiyao@126.com
  • 基金资助:
    国家重点研发计划(2021YFC2803302);国家自然科学基金(42176180);江苏省应急管理科技项目(YJGL-YF-2020-16);江苏省自然资源发展专项资金(海洋科技创新)(JSZRHYKJ202114)

Application of CatBoost model in water depth inversion

KONG Ruiyao1, XIE Tao1,2, MA Ming3, KONG Ruilin4   

  1. 1. School of Remote Sensing &Geomatics Engineering, Nanjing University of Information Science & Technology, Nanjing 210044, China;
    2. Laboratory for Regional Oceanography and Numerical Modeling, Qingdao National Laboratory for Marine Science and Technology, Qingdao 266237, China;
    3. Beijing Institute of Applied Meteorology, Beijing 100029, China;
    4. School of Software, Northwestern Polytechnical University, Xi'an 710129, China
  • Received:2022-01-25 Online:2022-07-25 Published:2022-07-28

摘要: 在多光谱遥感水深反演研究中,由于影响反演精度的因素较多,传统的水深反演模型具有一定局限性。机器学习算法在解决非线性高复杂问题上较有优势,将其应用在某些特定区域水深反演可提高反演精度。本文利用Sentinel-2多光谱遥感影像和LiDAR测深数据,以瓦胡岛为研究区域,构建CatBoost水深反演模型,与传统水深反演模型及Boosting中的XGBoost和LightGBM模型的反演精度进行比较。试验结果表明,经过参数优化后的CatBoost水深反演模型的决定系数、均方根误差、平均绝对误差和平均相对误差分别为96.19%、1.09 m、0.77 m和9.61%,准确性最高,效果更佳。

关键词: 水深反演, 多光谱遥感, Sentinel-2, 机器学习, CatBoost模型

Abstract: In multispectral remote sensing water depth inversion research, the traditional water depth inversion models have some limitations due to many factors affecting the accuracy of water depth inversion. Machine learning algorithms are more advantageous in solving nonlinear and highly complex problems, and their application in some specific areas of water depth inversion can improve the inversion accuracy. In this paper, using Sentinel-2 multispectral remote sensing images and LiDAR bathymetry data, it constructs CatBoost water depth inversion model with Oahu as the study area and compares the inversion accuracy with traditional water depth inversion models as well as XGBoost and LightGBM models in Boosting. It's showed in the experimental results that R-Square, root mean square error, mean absolute error, and mean relative error of the tuned CatBoost water depth inversion model are 96.19%, 1.09 m, 0.77 m and 9.61%, and the accuracy of the model is the highest, and the effect is more better.

Key words: water depth inversion, multispectral remote sensing, Sentinel-2, machine learning, CatBoost model

中图分类号: