基于高分四号卫星的北京市城区森林叶面积指数反演

Forest Leaf Area Index Inversion in Beijing Urban Area Using GF-4 Satellite Data

为分析太阳多角度的光谱信息在反演森林LAI中的应用潜力,本文基于高分卫星数据构建七种常用的植被指数,并添加纹理信息,对北京市开展区域LAI反演研究。采用随机森林和线性回归,通过提取不同太阳角度的高分卫星数据的植被指数以及影像纹理均值,与实测数据建立回归模型反演叶面积指数,R2和RMSE作为指标对模型进行精度评价。与以往的反演方法相区别,该文将太阳多角度光谱信息与纹理结合进行森林LAI的反演实验,并对反演算法进行敏感性分析。研究发现,基于高分卫星数据构建的单个角度下的植被指数与纹理数据的结合反演LAI的模型精度在0.3~0.5左右,添加太阳多角度后的植被指数和纹理信息的结合在一定范围内提高了北京市城区森林的LAI的反演精度,R2达到0.58,说明太阳多角度的光谱信息在反演森林LAI中具有一定的潜力,在定量遥感方面高分四号卫星理论秒级的时间分辨率具有一定开发应用的价值。本结果在该实验区验证可行,在其他地形区域的应用效果还需要进一步地研究。

In order to analyze the potential application of the solar multi-angular spectral information in inversion of forest LAI,the study constructs seven commonly used vegetation indexes based on GF-4 satellite data,and adds texture information to model the forest LAI in Urban Area of Beijing and conduct regional leaf area index(LAI)inversion research.The regression model for the inversion of LAI was established using random forest and linear regression,by extracting the vegetation index and image average texture of GF-4 satellite data from solar multi-angular,and the accuracy of evaluation of the model is evaluated with R2 and RMSE.In contrast to the previous inversion methods,this paper combines the solar multi-anglular spectral information with the texture information to carry out the inversion experiment of forest LAI,as well as the sensitivity analysis.Results show that the combination of vegetation indices and texture data at a single angle based on GF-4 satellite data is with an R2 of about 0.3~0.5;while by adding the solar multi-angle data improves the inversion R2 to 0.58.It can be concluded that that the solar multi-angle spectral information has some potential in the inversion forest LAI,and the high temporal resolution advantage of the GF-4 satellite is of great value in quantitative remote sensing.Despite that the experimental results are feasible in our experimental area,further research is still needed.