Desenvolvimento de um sistema hierárquico de classificação com aplicação baseada em abordagem fuzzy

Abstract

Land cover mappings represent important tools for the regional planning. However, the current mappings are related to very specific purposes and, consequently, they are limited in their capacity to define the wide variety of existing types of land cover. In that context, this paper aims at developing a wide and including hierarchical classification system for land cover mapping in regional scale, which should contribute for a future standardization of classes. Besides, it is intended to test that system for a study case that contemplates the use of a classification method based on fuzzy approach, which has shown to be more appropriate than conventional approaches. Therefore, it was proposed a hierarchical classification system with three detailing levels and a study case was defined with the specification of the test area and of the classification project. Then, the georreferencing of a TM/Landsat-5 image that comprises the test area was carried out. Later, it was applied a fuzzy classification approach in the TM/Landsat-5 image, starting from images of probability for the mapped classes and an uncertainty image were generated. Finally, it was produced a conventional output that represents the thematic mapping of the test area.