Use of SAR images for classification of brazilian forest formations

Abstract

Brazil has a large territorial area with a large cover of vegetation and several forest typologies with different physiognomies. It is necessary to map the forest areas in the country in order to know the spatial distribution and the dynamics of each forest formation. An efficient and reliable way to obtain this information is using remote sensing techniques, and radar - SAR - imaging can be applied, which in turn has been the focus of many researchers. Thus, the objective of the present study is to gather scientific productions related to the use of radar images applied to the mapping of different forests in Brazil, analyzing the most recent approaches and classification techniques. There was a significant application of SAR images in forests of the Amazon biome, mainly for the detection of deforestation. The images of the ALOS/PALSAR L-band radar system were the most used in the mapping of forest typologies, associated to several classifier algorithms, such as: Iterated Conditional Modes, Maximum Likelihood and random forest. The data types worked in the classifications varied according to the polarimetric capacity of each image, with emphasis on the greater use of backscattering coefficients and attributes extracted from matrix decompositions. It was also observed that most studies related SAR data to those obtained by optical sensors. Therefore, the present study made it possible to gather several applications of classification techniques for the discrimination of forest formations in Brazil using microwave imaging, indicating the potentiality of the various classifiers with SAR images, and showing that radar systems are an important technology that is being used for mapping forests in the country.

Keywords:
Remote sensing of vegetation; Synthetic aperture radar; Classification algorithm