Visualização de matrizes de covariância complexas: uma aplicação em dados PolSAR

Abstract

The monitoring of our planet through image sensors can be used to inspect: deforestation and desertification of forests; the water cycle; urban growth, among others. Among the technologies of remote sensing, PolSAR (Polarimetric Synthetic Aperture Radar) has been highlighting. The PolSAR enables remote sensing in almost all weather conditions. Each resolution cell in the PolSAR image is associated with a complex scattering matrix. One way to characterize the single-look PolSAR data is to use the complex covariance matrix. The covariance matrix is of extreme importance inMultivariate Analysis because it allows measure and evaluate the degree of dependence between the variables that compose the data set. This work presents a multivariate data visualization approach to visualize information contained in complex covariance matrices. The information aboutwhat type of homogeneous region is contained in the PolSAR image to be studied can be obtained by analyzing the data visualization approach provided in this work. For the validation of the approach, we applied our tool in non-simulated PolSAR images based homogeneous samples, i. e., samples that follow a certain of probability distribution. After that, the proposal was applied in three distinct targets, namely: a desertic region, a urban region and an aquatic region. Therefore, the implemented visualization approach in the R permits to discriminate the targets.