基于Sentinel-1结合Sentinel-2和Landsat8影像的水体提取方法比较——以鲁班水库为例

Comparison of water body extraction methods based on Sentinel-1 data, combined with Sentinel-2 data and Landsat8 images: taking Luban reservoir as an example

摘要: 基于sentinel-1数据结合sentinel-2数据和Landsat8数据，提取2017年—2021年鲁班水库五年间的水体年间变化，对所提取的水体面积进行分析，比较两种影像对鲁班水库水体提取的差异，经过水体指数的计算以及分析得到鲁班水库的水体面积。sentinel-2影像数据对水体提取的效果较好于Landsat8数据，sentinel-2影像总体分类精度达0.993，kappa系数为0.991，而Landsat8的总体精度为0.989，kappa系数为0.986。通过对鲁班水库这种大中型水体的水体提取选择使用sentinel-2数据较好。

Abstract: Base on sentinel-1 data, combined with sentinel-2 data and Landsat8 data, the annual changes of water bodies in Luban reservoir from 2017 to 2021 were extracted, the extracted water body area analysis was analyzed, and the difference of water bodies extracted by the two images of Luban reservoir were compared. The water body area of Luban reservoir in this study area was obtained through the calculation and analysis of the water body index. It was better to extract water from sentinel-2 image data than Landsat8 data. The overall classification accuracy of sentinel-2 image was 0.993 and the kappa coefficient was 0.991, while the overall accuracy of Landsat8 was 0.989 and the kappa coefficient was 0.986. Sentinel-2 data is more suitable for water body extraction in Luban reservoir, a large and medium-sized water body.