Abstract:
The article is devoted to the effectiveness research of controlled spectral-spatial classification of hyperspectral data methods in distinguishing vegetation types (agricultural cropes). A number
of approaches to increasing the classification accuracy by considering the pixels’ vicinity on different stages of data processing have been compared using the example of the test image fragment,
which was taken during the AVIRIS program. It was shown that method combining spatial processing of initial images with postprocessing of generated classification maps renders to be the most
effective.
Keywords:remote sensing, hyperspectral images, surface type classification, spectral and spatial features.
