基于Google Earth Engine的土地覆盖分类方法研究

Research on Land Cover Classification Method Based on Google Earth Engine

基于谷歌云在线处理平台(Google Earth Engine,GEE),以2019年全年Landsat-8 OLI影像为数据源,选取黑龙江省哈尔滨市为试验区,基于决策树算法、随机森林算法、最邻近算法、支持向量机算法开展土地覆盖分类方法研究,并对不同分类器方法提取结果进行精度评价。试验结果表明,基于GEE平台可以快速完成土地覆盖分类,各分类方法都可以达到较好的效果;采用决策树算法提取结果最优,总体精度为97.53%,Kappa系数为0.9586。

Based on the Google Earth engine(GEE),this paper takes Landsat-8 in 2019 as an example using OLI image as the data source,Harbin City of Heilongjiang Province as the experimental area,land cover classification methods are studied based on decision tree algorithm,random forest algorithm,nearest neighbor algorithm and support vector machine algorithm,and the accuracy of different classification algorithms is evaluated.The experimental results show that the land cover classification can be completed quickly based on the GEE platform,and each classification method can achieve good results;the decision tree algorithm has the best extraction result,the overall accuracy is 97.53%,and the kappa coefficient is 0.9586.