基于Sentinel-2影像的火烧迹地快速提取方法比较研究

利用Sentinel卫星数据，特别是其中对植被敏感的红边波段与短红外波段，针对提取灾后火烧迹地研究不足的问题，选取四川省冕宁县4月20日森林火灾发生前后的Sentinel-2卫星数据，使用不同的提取方法探究识别火烧迹地的潜力，并进行对比研究。实验结果表明，决策树分类法识别火烧迹地的能力最好，提取精度最高，BAI指数次之；其余方法均受道路、房屋和裸体不同程度的影响；用红边波段替换可见光红波段的效果相比原有指数提取效果并无明显提升。研究证明在各类方法中，采用决策树分类法能快速高精度地将火烧迹地准确提取出来。

Comparative Study of Rapid Extraction Methods of Burned Area Based on Sentinel-2 Images

Aiming at the problem of insufficient research on extraction of burned area, Sentinel satellite data were used, especially the red-edge band and short-infrared band that were sensitive to vegetation, and Sentinel-2 satellite data in Mianning County, Sichuan Province before and after the forest fire on April 20 were selected to explore the potential for identifying burned areas using different extraction methods, and a comparative study was conducted. The experimental results showed that the decision tree method had the best ability to identify burned areas, with the highest extraction accuracy, followed by the BAI index; the other methods were influenced to varying degrees by roads, houses, and bare land; the effect of replacing the visible light red band with the red-edge band was not significantly improved compared to the original index extraction effect. This study proved that the decision tree method could meet the requirements of high-precision and rapid extraction of burned areas.