海洋测绘Issue(4)：10-14,29,6.DOI:10.3969/j.issn.1671-3044.2025.04.003
多光谱信息改进的DBSCAN测深点云滤波方法
Multispectral information-assisted filtering of bathymetric point clouds
摘要
Abstract
Noise in ICESat-2 photon measurement data significantly interferes with the extraction of seabed topography,making an effective filtering algorithm crucial for the validity of ICESat-2 bathymetric data.Addressing the limitations of the traditional DBSCAN filtering algorithm,this study proposes an improved DBSCAN filtering approach incorporating multispectral imagery information,based on the principle that both ICESat-2 bathymetric point clouds and multispectral band ratios are influenced by water depth variations.The proposed algorithm employs M-estimation to fit the residual points after outlier removal and applies a confidence interval-based filtering process to refine the residual points,thereby enhancing the precision of ICESat-2 data processing.Comparative analyses with in situ measurements in the waters around Maui Island and Ganquan Island demonstrate,the proposed method yields improvements in both MAE and RMSE,significantly enhancing the accuracy of ICESat-2-derived bathymetry.This approach provides robust technical support for shallow-water depth retrieval.关键词
水深探测/ICESat-2测深/多光谱影像/密度聚类算法/点云滤波Key words
bathymetry/ICESat-2 bathymetric/multispectral imagery/DBSCAN/point cloud filtering分类
天文与地球科学引用本文复制引用
张琪,张帅,李程,张凯,陈卓..多光谱信息改进的DBSCAN测深点云滤波方法[J].海洋测绘,2025,(4):10-14,29,6.基金项目
国家重点研发计划(2022YFC2806600 （）
2022YFC2806602). （）
