Multispectral Image Analysis for Agricultural Crop Classification in Peru Using Deep Learning Techniques

Flores A.; Villarroel J. · Proceedings - 2025 2nd IEEE Afro-Mediterranean Conference on Artificial Intelligence, AMCAI 2025 (2025) · DOI: 10.1109/amcai66110.2025.11474318

Resumen

Crop classification is crucial for agricultural management, especially in the context of precision agriculture in Peru. This study presents the results of three investigations that utilize remote sensing and deep learning technologies to classify agricultural crops across different phenological phases. Using multispectral images obtained from UAVs and satellites, the study focuses on classifying table grapes, corn and quinoa crops, three key agro-exportable products in Peru. The research emphasizes the use of semantic segmentation networks in conjunction with vegetation indices such as NDVI to improve the accuracy of crop classification. The results demonstrate that deep learning techniques outperform traditional classification methods in terms of accuracy and efficiency, particularly when integrating various image types and addressing phenological variations. This work aims to optimize crop management and resource use while contributing to food security through better decision-making based on geospatial data.

Instituciones: Pontificia Universidad Católica del Perú