Agricultural property monitoring using geospatial foundation models.
No hay miniatura disponible
Fecha
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Resumen
Descripción
In this paper we present a preliminary study on the application of geospatial foundation models to the satellite-based monitoring of agricultural properties. Images from the Sentinel-2 mission are processed with the Clay foundation model in order to produce embeddings in the model’s latent space, which are later reduced in dimension with a principal component analysis and plotted in 2 dimensions to enable visual inspection. A case study was carried out considering scenes from a property located in the Guia Lopes da Laguna agrotechnological district, a partner of Semear Digital. Two management situations were taken into account: a) traditional pasture; and b) initial preparation for an integrated livestock-forest system. Differences in the embeddings from both scenarios were analyzed, and early observations indicate that the method has potential and may become a consolidated practice in the area.
Organização: Silvia Maria Fonseca Silveira Massruhá, Durval Dourado Neto, Luciana Alvim Santos Romani, Jayme Garcia Arnal Barbedo, Édson Luis Bolfe, Ivan Bergier, Maria Angelica de Andrade Leite, Vitor Del Alamo Guarda, Catarina Barbosa Careta.
Organização: Silvia Maria Fonseca Silveira Massruhá, Durval Dourado Neto, Luciana Alvim Santos Romani, Jayme Garcia Arnal Barbedo, Édson Luis Bolfe, Ivan Bergier, Maria Angelica de Andrade Leite, Vitor Del Alamo Guarda, Catarina Barbosa Careta.
Palabras clave
Aprendizado de máquina, Aprendizado profundo, Integração pecuária-floresta, Projeto Semear Digital, Deep machine learning, Livestock-forest integration, Sensoriamento Remoto, Remote sensing