Algoritmo de inteligencia artificial para la detección de cultivos de cacao (Teobroma cacao L.), banano (Musa paradisiaca L.) y palma africana (Elaeis guineensis J.).

Abstract

It analyzes the expansion and intensification of unregulated and unsustainable agriculture, mainly in Latin America, focusing on Ecuador, where land degradation is closely related to unsustainable agricultural practices. It highlights the importance of comprehensive land-use planning and the use of remote sensing to assess changes in land use and land cover. The research focuses on the northern area of the provinces of Guayas and Los Rios in Ecuador, focusing on the processing of satellite images to detect banana, cocoa and palm crops. It uses an experimental and quantitative methodology to compare the effectiveness of different classification methodologies, evaluating the accuracy of the artificial intelligence method Random Forests using Sentinel-2 satellite images. The results of the study include preprocessing of satellite images, calculation of NDVI and RESI spectral indices, and supervised classification of Sentinel-2 images using the Maximum Likelihood and Random Forests methods. Detailed statistical analyses were performed, including validation of results using confusion matrices and calculation of the Kappa coefficient. In conclusion, the manuscript provides an in-depth understanding of how remote sensing and artificial intelligence can help in the detection and classification of crops in specific areas, contributing to agricultural management and sustainable land planning.