基于高光谱影像的高寒牧区土地覆盖分类与草地生物量监测模型

Land cover classification and grassland biomass monitoring model in alpine pastoral area based on HJ-1Ahyperspectral image

利用环境减灾卫星HJ-1A高光谱图像数据,分析了研究区不同土地覆盖类型的波谱曲线特征,比较了监督分类和光谱角分类方法对高光谱影像的分类精度,研究了高寒牧区草地生物量超光谱遥感监测模型。结果表明,1)不同地物波谱曲线的吸收位置和吸收深度等波谱特征在可见光波段具有较大差异,在近红外波段吸收特征相似。在可见光波段,云和植被的吸收位置最少,都只有1处,但云的吸收深度小于植被;裸地吸收位置有5处;水域吸收位置最多,有6处。2)光谱角与监督分类均适于高光谱影像分类,但光谱角分类方法的总精度可达85.9%,远高于监督分类法。3)依据草地生物量与9种植被指数间的回归分析结果,选出了适合研究区草地植被生物量动态监测的两种植被指数,即归一化植被指数和比值植被指数。

Using hyperspectral image of HJ-1A satellite,the standard spectral curves of different objects in the study area were analyzed;land cover classification accuracy between supervised classification and spectral angle mapping method were compared;and the grassland biomass monitoring models based on hyperspectral remote sensing in alpine pastoral area were also studied.The results show that：1） the absorption position,absorption depth and other spectral characteristics of the standard spectral curves of different objects are different in the visual bands,but they are similar in near-infrared bands.In the visual bands,there is only one absorption position in the standard spectral curves of clouds and vegetation,and the absorption depth of clouds is less than that of vegetation.In the standard spectral curves of bare lands and water areas,there are five and six absorption positions,respectively.2） Both the spectral angle mapping and supervised classification methods are suitable for hyperspectral image classification.The overall classification accuracy of spectral angle mapping method reaches to 85.9% and is much higher than that of supervised classification approach.The spectral angle mapping method can recognize the objects under thin clouds,cloud shadows and mountain shadows.3） Based on regression analyzed results between grassland biomass and 9 vegetation indices,two vegetation indices of Normalized Difference Vegetation Index（NDVI） and Simple Ratio Index（SR） are suitable for grassland biomass monitoring in the study area.HJ-1A hyperspectral data has been successfully applied in classification of land cover of alpine pastoral areas.This research laid foundations of further studying in HJ-1A hyperspectral imaging data.