Author(s): YANO, I. H.; CARVALHO, M. L. de M; MARCHIORI, L. F. S.; SILVA, F. C. da

Summary: Productivity estimates play a crucial role for sugarcane producers and sugar mills in planning production, aligning it with demand forecasts. Manual estimations demand considerable effort and time, prompting exploration into alternative productivity estimation methods such as aerial imaging using drones. Within imaging techniques, productivity estimation occurs indirectly through the analysis of vegetation indices. The widely recognized vegetation index, NDVI, necessitates costly near-infrared (NIR) cameras, making it inaccessible to many producers. Our approach utilized drone imagery captured by more affordable RGB cameras, which are feasible for a larger number of producers. We applied six regression models alongside a stacking model that amalgamated these six models for estimating sugarcane production using the eight RGB vegetation indices. Initial tests revealed a Mean Absolute Percentage Error (MAPE) of less than 13%. This level of accuracy is considered favorable when benchmarked against similar studies and presents encouraging prospects for future research.

Publication year: 2024

Types of publication: Paper in annals and proceedings

Unit: Embrapa Digital Agriculture

Keywords: Aprendizado de máquina, Cana de Açúcar, Drone, Machine learning, Rendimento, Sugarcane, Unmanned aerial vehicles, Yields