论文部分内容阅读
本研究以河南省嵩县为研究区域,构建了地形因子、植被指数和纹理特征等分析研究区优势树种在特定地形条件下与植被指数特征和纹理特征的差异形成的这些树种的光谱差异;利用基于知识的决策树分类方法和支持向量机分类这两种方法,进行各优势树种的分类识别,研究和探索出一种适合伏牛山地区地形的条件树种识别方法,得到如下结论:(1)在嵩县山区地貌条件下的树种分类识别:地形因子起着不可或缺的作用。(2)利用植被指数和纹理特征的支持向量机分类比利用植被指数的支持向量机分类Kappa系数提高了0.0906,总精度提高了4.63%;利用植被指数、纹理特征和地形因子的支持向量机分类比利用植被指数和纹理特征的支持向量机分类Kappa系数提高了0.1035,总精度提高了7.74%。(3)对于伏牛山嵩县地形条件下的遥感影像树种分类识别,利用纹理特征、植被指数、光谱信息和地形因子的支持向量机是最佳的分类方法,满足大尺度遥感监测。
In this study, Songxian County, Henan Province as the research area, constructed the topographic factors, vegetation index and texture features analysis of the dominant tree species in the study area under the specific topography and vegetation index characteristics and texture characteristics of the difference between the formation of these species spectral differences; Based on the knowledge-based decision tree classification method and support vector machine classification method, the classification and identification of each dominant tree species were studied and a conditional tree species identification method suitable for the topography in Funiu Mountain region was studied. The conclusions are as follows: (1) Songshan County landforms under the classification of tree species identification: the terrain factor plays an indispensable role. (2) SVM classification using vegetation index and texture features Compared with the SVM classifier with vegetation index, the Kappa coefficient improved by 0.0906 and the overall accuracy increased by 4.63%. Using the SVM classification of vegetation index, texture features and topographic factors Compared with the SVM classifier using vegetation index and texture features, the Kappa coefficient increased by 0.1035 and the overall accuracy increased by 7.74%. (3) For the classification and recognition of remote sensing image tree under the topographic conditions in Songshan County of Funiu Mountain, the SVM based on texture features, vegetation index, spectral information and topographic factors is the best classification method to meet the large-scale remote sensing monitoring.