基于GEE和Landsat时间序列数据的香格里拉森林类型分类研究

Forest Types Classification of Shangri-La Based on Google Earth Engine and Landsat Time-series Data

基于Google Earth Engine云平台和2014—2017年Landsat OLI影像序列,根据其在时间域上的光谱特征,结合植被指数特征、地形和温度特征,采用随机森林分类算法,开展香格里拉森林类型分类研究。结果表明:不同森林类型的生长轨迹有明显差异,4种森林类型在冬季的植被指数差异最明显;时间序列影像数据能够提供不同森林类型的物候差异特征,弥补单一日期影像难以区分不同森林类型的困难;研究区森林/非森林覆盖的总体精度为97.17%,Kappa系数为0.943,森林类型分类的总体精度87.78%,Kappa系数为0.80。基于Landsat时间序列的方法能够提供一个精度较高的森林分类产品,可为基于森林类型制图的应用提供帮助。

Based on the Google Earth Engine platform and Landsat OLI time-series data from 2014 to 2017,combined with vegetation index,topography and temperature related features,the forest type classification of Shangri-La was studied by using random forest classification classifier.The findings indicates that the different types have obvious phenological characteristics difference in various seasons,among which 4 forest types were the most in winter.Therefore,the time-series data can provide the phenological difference characteristics for classification,thus benefitting the classification of forest types.Accuracy assessment of forest/non forest classification indicates that the overall accuracy and Kappa coefficient are 97.17%and 0.943,respectively.And the forest type classification’s overall accuracy and Kappa coefficient are 87.78%and 0.80,respectively.The method can provide a forest classification product with high precision and is helpful for mapping forest types.