基于GEE的徐州市土地利用分类研究

Land use classification of Xuzhou based on GEE

及时、准确地获取土地利用信息,可为城市发展和生态环境保护提供参考依据.基于谷歌地球引擎(Google earth engine)云平台,联合Sentinel-1 SAR数据、Sentinel-2 MSI高分辨率光学影像数据、SRTM高程数据等多源数据构建分类特征集,利用随机森林算法对徐州市2021年的土地利用类型进行分类,并对分类结果进行精度评价.研究结果表明:1)地物的光谱特征,尤其是归一化水体指数(NDWI)对分类结果贡献较大,多特征集的分类精度明显高于单一光谱特征;2)综合利用地物光谱特征、纹理特征、地形特征和雷达后向散射特征进行随机森林分类,分类精度最高,达93.55%,水体和耕地的分类效果明显优于裸地和建设用地;3)徐州市的主要土地利用类型为耕地和建设用地,两者面积占比达92.44%,水体主要分布于铜山区和新沂市,林草地和裸地分布较少.

Land use information acquired timely and accurately can provide reference for the city's sustainable development and ecological environment protection.Based on the platform of Google earth engine(GEE),Sentinel-1 SAR,Sentinel-2 MSI and SRTM were combined to construct the classification feature set.The random forest algorithm was used to classify the land use in Xuzhou in 2021,and the accuracy of the classification results was evaluated.The results showed that the spectral features,especially the NDWI contributed significantly to the classification results,and the accuracy of multi-feature sets was higher than that of single spectral features.Spectral feature,texture feature,terrain feature and radar backscattering feature were used for random forest classification,with the highest accuracy reaching 93.55%,the classification accuracy of water body and cultivated land was significantly better than that of bare land and construction land.The main land use in Xuzhou was cultivated land and construction land,accounting for 92.44% of the total area,the water body was mainly distributed in Tongshan district and Xinyi city,with less forest and grass land and bare land.