Automated classification of citrus areas using SVM and NDVI temporal patterns: applications for precision agriculture and logistics management

• Autores: Sánchez-Galván, Fabiola, García-Rodríguez, Rogelio, Salas Martínez, Paulino, Altamirano Herrera, María Xóchitl, Bautista-Santos, Horacio
• Formato: artículo
• Fecha de Publicación: 2025
• Institución: Universidad de Lima
• Repositorio: Revistas - Universidad de Lima
• Lenguaje: español
• OAI Identifier: oai:revistas.ulima.edu.pe:article/7858
• Enlace del recurso: https://revistas.ulima.edu.pe/index.php/Interfases/article/view/7858
• Nivel de acceso: acceso abierto
• Materia: sustainable agricultural supply chains; precision agriculture; agrologistics; cadenas de suministro agrícolas sostenibles; agricultura de precisión; agrologística

This study developed a model based on Support Vector Machines (SVM) and Normalized Difference Vegetation Index (NDVI) time series to classify citrus areas in Álamo, Veracruz, Mexico. MODIS images (MOD13Q1, 250 m resolution) from 2003 to 2022 were used, processed using radiometric correction, noise filtering, and temporal harmonization. Training areas were classified into four categories: citrus, natural vegetation, grasslands, and urban areas, using 3,759 time series, 50 % of which were positive for citrus. The SVM model (RBF kernel: γ = 0.1, C = 10) achieved an accuracy of 91.4 % using 5-fold cross-validation, with 88% success in citrus and 93.9 % in non-citrus samples. The results showed an average NDVI of 0.74 for citrus, distinguishable from weeds (0.87), although with challenges in small plots due to spatial resolution. The estimates coincided with official data (SIACON) in 2021 (548 ha difference), although they presented discrepancies in years with droughts (2007, 2015) or management changes (2019-2020). Climatic and anthropogenic factors were identified as affecting NDVI dynamics, demonstrating the model’s usefulness for agricultural monitoring limitations include pixel mixing in heterogeneous areas. This work demonstrates that the SVM-NDVI approach is robust for classifying citrus areas at a regional scale, with potential applications in logistics management, such as transport route optimization and crop planning using spatiotemporal NDVI patterns. These findings open up opportunities to integrate remote sensing and machine learning into sustainable agricultural supply chains.