MAPEAMENTO DE ÁREAS PERMEÁVEIS, UTILIZANDO IMAGENS ORBITAIS E APRENDIZADO DE MÁQUINA

Abstract

Permeable areas play a crucial role as indicators of a city's environmental quality, providing ecosystem services that benefit both human well-being and ecological balance. Likewise, permeable areas play a fundamental role in the hydrological characteristics of cities. Mapping these urban characteristics is essential to support urban planning and management. Therefore, the objective of this work was to use remote sensing data processed by machine learning to map the permeable areas in the Imbirussu stream basin, located in Campo Grande, Mato Grosso do Sul, Brazil. To conduct the study, we employed high-resolution satellite imagery and a supervised object-based classification approach using the Random Forest machine learning algorithm. These experiments were conducted with images from Google Earth (GE) Pro dated April 20, 2023 and the ESRI ArcGIS Pro 3.0 software. The results obtained for permeable areas, we achieved an overall accuracy of 92.63% and an F1-score of 92.43%. These results indicate that the strategy adopted is effective in the task of mapping permeable areas, demonstrating the potential for using data from GE Pro, whose historical images are generally available for all cities in the world. This study represents a breakthrough in integrating machine learning analyzes with high-resolution imaging data. This allows urban decision-makers to base their choices on concrete information, resulting in more effective and sustainable policies for the development of cities.