基于GEE云平台的山东省不透水面提取

Extraction of Impervious Surface in Shandong Province Based on Google Earth Engine and Sentinel Data

不透水面是衡量城市化发展水平的重要指标,及时、准确地掌握城市不透水面的动态变化对于监测城市地区的发展和环境变化至关重要。利用Google Earth Engine(GEE)云平台和Sentinel-1SAR数据、Sentinel-2光学数据以及随机森林算法对山东省2020年4月不透水面进行提取。结果表明:使用光学和SAR数据融合的山东省不透水面提取精度为92.17%,Kappa系数为0.79;雷达特征的重要性优于光学特征,光学与雷达数据结合提取不透水面精度更高,错提明显减少;雷达特征对于面状不透水面的提取效果优于线状不透水面,实际使用中应针对不同的应用选择合适的特征。

The impervious surface is an important indicator to measure the level of urbanization development,and timely and accurate grasp of the dynamic changes of the urban impervious surface is essential for monitoring the development of urban areas and environmental changes.This paper uses Google Earth Engine(GEE)cloud platform and Sentinel-1 SAR data,Sentinel-2 optical data and random forest algorithm to extract impervious surfaces in Shandong province in April 2020.The results show that,the extraction accuracy of impervious surface in Shandong province using the fusion of optical and SAR data is 92.17%,and the Kappa coefficient is 0.79.The importance of radar features is better than that of optical features.The combination of optical and radar data extracts impervious surfaces with higher accuracy and significantly reduces misclassification.The extraction effect of radar features on planar impervious surfaces is better than linear impervious surfaces.In actual use,suitable features should be selected for different applications.