Mapping land use cover in Amazon region with COSMO SkyMed data

The Amazon region has intense clouds coverage and the use of radar images becomes an option for land use and land cover mapping to better characterize the deforestation areas already mapped by PRODES Project. This work analyses the potential of COSMO-SkyMed image to discriminate four classes of land use. It has used the original and filtered (Frost and Lee filters) data and the higher confusion occurred between woody vegetation and undergrowth classes. Among the 20 classifications, the results with the filtered image by Lee filter with window of 7×7 pixels presented a better performance, but other techniques should be evaluated.