极端干旱区荒漠稀疏河岸林遥感分类研究

研究以位于极端干旱区的塔里木河干流中下游地区为例,基于Landsat TM影像,结合决策树分类、几何光学模型与光谱角匹配,解决混合像元信息分解,实现干旱区稀疏荒漠河岸林类别识别。首先从遥感视角的角度,将地物分解为目标和背景,提出塔里木河干流荒漠河岸林植被分类系统;其次以多变量决策树法将非荒漠植被信息剔除,采用几何光学模型模拟各类荒漠植被的像元光谱,最后以光谱角匹配的方法将荒漠植被进一步进行分解,得到塔里木河干流中下游地区典型研究区的植被分类专题图,分类精度结果表明:基于混合像元分解与几何光学模型的分类方法总精度达到了79.43%,Kappa系数为0.718,表明分类质量良好。

Classification of Sparse Desert Riparian Forest in Extreme Arid Region

Taking the desert riparian forest belts along both riversides of the middle and lower reaches at the Tarim River Basin as the research object and making use of Landsat TM data, a new classify method of combining decision tree, Geometric Optical models and Spectral Angle Mapper is introduced to identify the desert riparian forest sort in extreme arid region. Firstly, a new classification system of desert riparian forest was brought forward, dividing the target into object and background from view of remote sensing. Secondly, the non-desert vegetation information was masked off by using the method of decision tree;the spectrum of the desert riparian forest pixels were simulated with the pure Geometric Optical and Geometric Optical-Radiative Transfer model, then to map the vegetation of the study area using Spectral Angle Mapper based on the pixel spectrum simulated. The results indicate that the quality of classification is good, with the accuracy coefficient to 79.43% and the Kappa coefficient to 0.718.