Aplicação de sensoriamento remoto na delimitação de áreas úmidas: análise dos métodos empregados

Abstract

Wetlands are ecosystems of extreme importance, both nationally and internationally, but many of them have not yet been properly delineated, which hinders their monitoring and protection. Over the years, there have been several developments in the field of remote sensing and the application of this practice in conjunction with the use of artificial intelligence (machine learning, deep learning), allowing mappings to be carried out remotely and efficiently. Combining these factors, the delineation of wetlands has been conducted in research using different methods. This Systematic Literature Review aimed to compare these methodologies and define which would be the most suitable for the southwestern region of São Paulo, in addition to discussing the role of artificial intelligence in these processes. From this, it was determined that the best method would be the application of a hybrid workflow, combining multisensor data, deep learning, and object-based segmentation, being capable of mapping marshes, wet fields, and palm swamps in the region, as well as highlighting that the role of artificial intelligence is to assist in the process, automating stages and reducing the time required for mapping.