Vegetation classification technology of hyperspectral remote sensing based on spatial information

The vegetation classification effects of traditional remote sensing methods only considering spectral information are unsatisfactory because the hyperspectral data usually have noises, so it is particularly important to blend spatial information for vegetation classification. Firstly, vegetation information was extracted by NDVI threshold value, and minimum noise fraction (MNF) was used to compress Hyperion hyperspectral images, and the first 60 components were selected. And then a kind of hyperspectral image vegetation classification method with the combination of spatial and spectral information was applied to complete the vegetation classification in study area. The results indicated that the average classification accuracy of all vegetation types was 90.3%, while the average classification accuracy of maximum likelihood method was only 70.0%. The vegetation classification method of hyperspectral remote sensing combining spatial information can effectively weaken the noises and to a certain extent improve classification accuracy, so the method proposed in this paper has certain reference value in actual application.