Mapeamento de fitofisionomias do Cerrado utilizando classificação orientada objetos (GEOBIA) e imagens CBERS 4A

Abstract

The mapping of vegetation cover and the trajectory of technological development in the construction of a Brazilian cartographic memory brings products at different levels: from the most detailed to the most generalized. Biomes that over the years have been represented with an increasingly low level of detail. The generalization of the Cerrado biome vegetation nowadays raises a question in comparison to maps produced without the use of data obtained by remote sensing. Therefore, the present work aims to analyze the representations generated from automated techniques of the Cerrado phytophysiognomies in relation to conventional mapping techniques, in the Ceres Microregion in Goiás. Thus, demonstrating that it is possible to collect data remotely through geoprocessing tools combined with remote sensing images. Images from the CBERS 4A WPM satellite were used to map land use and land cover. The method used was the object-oriented classification technique (GEOBIA), an approach that consists of the application of segmentation and classification based on decision rules. The application of this technique generated two thematic maps, the first representing the classes referring to the 1993 vegetation map proposed by the IBGE; and the second referring to the classes that Mapbiomas classifies in the year 2020. The validation of the maps was 0.8; excellent according to statistical parameters. The results obtained through the validation of these maps brought advantages over the conventional ways of representing phytophysiognomies. Advantages arising from the use of image processing techniques to automate some recognition patterns (shape, color and texture) that are used only in visual interpretation. The use of GEOBIA also resulted in the generation of a more detailed product in relation to the vegetation cover represented in the Mapbiomas platform.