Detecção de placas solares em imagens de satélite utilizando aprendizado não supervisionado

Abstract

This study aims to develop an automated approach for satellite image segmentation, focusing on detecting photovoltaic panels in the municipality of Aracaju - SE. The satellite images were provided by SEMFAZ through the Technical Cooperation Agreement between UFS and the Municipal Finance Department of Aracaju (SEMFAZ) No. 01/2022. The proposed methodology employs the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm to perform unsupervised image segmentation, enabling the identification of patterns related to photovoltaic structures under varying lighting conditions and noise. PostgreSQL was chosen as the database management system, utilizing the JSONB data type for efficiently storing extracted information. Additionally, the analysis of DBSCAN parameters, such as eps and minPts, proved crucial in optimizing the segmentation of image features. To facilitate user interaction, a software tool was developed to allow parameter adjustment, selection, and storage of the identified photovoltaic panel information. The results demonstrate that the proposed approach successfully achieved its objectives, enabling efficient segmentation and structured data storage. This study contributes to the mapping of solar energy structures and paves the way for the application of supervised methods as a robust dataset is built.