Análise do potencial de imagem TerraSAR-X para mapeamento temático no sudoeste da Amazônia Brasileira

Abstract

The objective of this work was to analyze the potential use of SAR polarimetric images from the TerraSAR-X sensor system, at StripMap mode, to map land use and land cover in SW Brazilian Amazon. Amplitude images at polarizations A HH, A VV, A, derived from the co-variance matrix, as well as the entropy A Entropia, derived from the decomposition of targets by eigenvalues, are parts of the datasets investigated individually or in combined form. Two classifiers were used: the first is based on statistical functions of maximum likelihood (MAXVER), and the second is the contextual method (Context). The thematic results from these classifications were evaluated by a confusion matrix and by the Kappa index. Summarizing we can state that the components A and A Entropia, gave a significant contribution to the image classification procedure, considering specially the Context method, whose performance reached 78% of Global Accuracy and a Kappa index of 0.70.