北京测绘2026，Vol.40Issue(8)：1049-1055,7.DOI:10.19580/j.cnki.1007-3000.2025120061
基于多源遥感数据的土地利用信息提取方法研究
Land use information extraction methods on multi-source remote sensing data
摘要
Abstract
To quantitatively evaluate the contribution of various features derived from multi-source remote sensing data to land use extraction and screen the optimal feature combination,this study built a composite feature set incorporating spectral bands,spectral indices,digital elevation model(DEM),and synthetic aperture radar(SAR)features using Sentinel-1 SAR data,Sentinel-2A optical data,and topographic data.Four classification algorithms,namely minimum distance classifica-tion,classification and regression tree(CART),support vector machine(SVM),and random forest(RF),were applied to conduct land use information extraction.The results show that when only spectral bands are used,the performance of the four classification methods varies,with RF achieving the highest Kappa coefficient of 0.85.After incorporating spectral indi-ces,classification accuracy improves to a certain extent,but the performance remains limited in distinguishing paddy fields from dry lands.When DEM and SAR data are further integrated,classification accuracy increases significantly,with Kappa coefficients reaching 0.90 and 0.92,respectively.Integrating multi-source remote sensing data and features with the random forest algorithm can effectively enhance the accuracy of land use information extraction in hilly regions of southern China.关键词
土地利用信息/随机森林/合成孔径雷达(SAR)/哨兵数据/多源数据Key words
land use information/random forest algorithm/synthetic aperture radar(SAR)/sentinel data/multi-source data分类
天文与地球科学引用本文复制引用
秦瑞,曹仕,卢霞,平李娜,彭寄清..基于多源遥感数据的土地利用信息提取方法研究[J].北京测绘,2026,40(8):1049-1055,7.基金项目
湖南省自然科学基金(2024JJ8381,2024JJ8373,2025JJ80064) （2024JJ8381,2024JJ8373,2025JJ80064）
