Potencial da Multipolarização e da Multifrequência de Imagens de Radar na Identificação da Cobertura da Terra na Amazônia Oriental

Potential of Multipolarization and Multifrequency Radar Images in Identification of Land Cover in the Eastern Amazon

Abstract

The objective of the present work is to analyze the potential of synthetic aperture radar images to identify the different classes of land cover representative for the Eastern Amazon, in a protected area. Images distributed free of charge were analyzed referring to the ALOS PALSAR-2 (L-band, HH and HV polarization) and Sentinel 1A (C-band, VV and VH polarizations) sensors, referring to the month of November 2022, which were evaluated separately and integrated for identification, through machine learning classification, of the following classes: Forest Cover, Grassland Cover, Wet Fields and Water Cover. The results show that the targets of interest, in a representative environment of the Amazon, are better differentiated through the multifrequency integration and multipolarization between the C and L bands, allowing to achieve excellent classification accuracy and all the chosen classes presented satisfactory Kappa index.