Uso de deep learning para identificação da arborização no contexto da região metropolitana do Rio de Janeiro

Abstract

Urban areas play a central role in climate change while also being strongly affected by its impacts. In the Rio de Janeiro Metropolitan Region (RMRJ), characterized by high population density and intense urban expansion, urban trees play an essential role in mitigating environmental effects and promoting quality of life. However, large-scale measurement of urban tree cover requires advanced methodologies. This study applies deep learning techniques to remote sensing data in order to identify and map urban tree cover in the RMRJ. The research involves: a review of the state of the art on the use of deep learning for vegetation detection in urban contexts; evaluation of model performance through accuracy metrics; and a quantitative analysis of vegetation distribution, considering spatial aspects. As a result, the model developed in this study achieved excellent performance, with high accuracy values in all evaluations, consistently exceeding 0.9 across all performance metrics.