Classification of closed vegetation types from the fusion of remote sensing images

Abstract

On-site monitoring of the Cerrado vegetation cover becomes impracticable due large coverage. The objective of this work was to perform the image fusion of the CBERS-4 satellite and to analyze the performance of the use of the fusion in the process of vegetation classification and soil occupation in the RPDS - Legado Verdes do Cerrado. The MUX sensors were selected to form the RGB false-color image and the PAN5m, followed by the substitution fusion process, using the IHS method. The results revealed that the fusion of images generated a gain of visual and spatial quality, improving the information used as a basis for classification. The thematic map had an overall performance of 84.83% and average confusion of 15.17% when using fusion, and 77.66% and 22.34% when using only image with RGB composition. On-site sampling contributed to the acquisition of the sample polygons and the correct definition of the classes, but the confounding values still considered high.