Fine Classification of Vegetation Under Complex Surface Cover Conditions with Hyperspectral and High-Spatial Resolution: A Case Study of the Xisha Area, Chongming District, Shanghai

Abstract Since both diversity and similarity exist among different vegetation types and since differences and similarities are reflected mainly in geometric morphology and in physical and chemical characteristics, the feedback signals of remote sensors can exhibit both similarities and differences. In vegetation remote sensing, the phenomenon of foreign bodies within the same spectrum is likely to occur. The complex mixed environment of land and water, which is a combination of cultivated land, wetland, woodland, and grassland areas, leads to the occurrence of such problems. In this case, relying on data with a high spatial resolution or high spectral resolution alone cannot effectively improve the accuracy of vegetation classification. Both hyperspectral and high spatial resolutions (H 2 , with both nanometre spectral resolution and submeter spatial resolution) can ensure that both morphological and spectral characteristics are considered when improving vegetation classification accuracy. In this paper, the Xisha wetland in Chongming and the surrounding cultivated land and forestland areas comprise the experimental area. H 2 images with both high spatial and high spectral resolutions were collected using an unmanned aerial vehicle platform. Forty-five plant cover types (and 10 non-vegetation features) were classified according to the spectral features of ground objects. The overall classification accuracy reached 97.86%, and the kappa coefficient reached 0.9725.