A Systematic Classification Method for Grassland Community Division Using China’s ZY1-02D Hyperspectral Observations

The main feature of grassland degradation is the change in vegetation community structure. Hyperspectral-based identification basis and a prerequisite for large-area high-precision monitoring management. To obtain distribution pattern communities Xilinhot, Inner Mongolia Autonomous Region, China, we propose systematic classification method (SCM) hyperspectral using China’s ZiYuan 1-02D (ZY1-02D) satellite. First, sample label data were selected from field-collected samples, map data, function zoning Nature Reserve. Second, spatial features images extracted extended morphological profiles (EMPs) based on reduced dimensionality principal component analysis (PCA). Then, they input into random forest (RF) classifier to preclassification results communities. Finally, reduce influence salt-and-pepper noise, similarity probability filter (LSPF) was used postclassification processing, RF again final results. showed that, compared with other seven (e.g., SVM, RF, 3D-CNN) methods, SCM obtained optimal an overall accuracy (OCA) 94.56%. In addition, mapping its ability accurately identify various ground objects large-scale scenes.