基于高光谱影像的树种分类

Tree Species Classification with Hyperspectral Image

应用环境与灾害监测预报小卫星(HJ-1A)的超光谱成像仪(HSI)高光谱影像,对大兴安岭地区塔河林区进行树种分类。HSI数据经过去条带和大气校正预处理,获取准确的树种光谱信息,分别采用波谱角填图法和线性波谱分离法进行树种分类。结果表明:在树种识别中,基于线性波谱分离的方法较基于波谱角填图法要优越,线性波谱分离方法的总体精度为72.0%,Kappa系数为0.600;波谱角填图法的总体精度为62.5%,Kappa系数为0.459。

We classified the tree species in Tahe County in Daxing＇ an Mountains with the HSI hyperspectral image acquired by the small satellite constellation of environmental and disaster monitoring and forecasting （HJ-1A）. The HSI data can only obtain precise tree species spectral information ＇after removing strips and correcting the atmospheric effect of remote sensing image. We classified the tree species by spectral angle mapping （SAM） method and linear spectral separation, respective- ly. Tile overall precision of spectral angle mapping was 62.5% with Kappa of 0.459, and that of linear spectral separation was 72.0% with Kappa of 0.60, indicating that the classification method based on linear spectral separation could get better result than spectral angle mapping in tree species classification.