AVALIAÇÃO DE ÍNDICES ESPECTRAIS NA CONTAGEM DE PLÂNTULAS DE MILHO ATRAVÉS DE SENSOR RGB DE BAIXO CUSTO

Abstract

Maize has become the main agricultural crop cultivated on a global scale, surpassing 1 billion tons in annual production. In addition to its large output, maize is highly versatile, serving as raw material for a wide range of products. In Brazil, the 2024 crop forecast is 119.6 million tons, according to the National Supply Company (Conab). In this context, remote sensing has gained prominence by collecting field data and enhancing timely decision-making. RGB sensors, despite their low cost and lower spectral resolution, are advantageous due to their cost-effectiveness. The objective of this study was to evaluate seedling counting using vegetation indices calculated from the bands of the electromagnetic spectrum within the visible range, captured by low-cost sensors mounted on unmanned aerial vehicles (UAVs). The study was conducted in an experimental maize field of approximately 0.2 hectares located in the municipality of Rio Verde, state of Goiás, Brazil, at the V3 phenological stage during the 2024 growing season. The experimental area was surveyed by UAV, followed by image processing, vegetation index calculation, manual seedling counting, and image-based prediction. Statistical analysis revealed that the GLI (Green Leaf Index) was the most effective predictor, showing a significant positive correlation with the seedling counts performed in the field. This study paves the way for future research integrating more advanced techniques for plant detection and other agricultural targets, thus contributing to the advancement of digital agriculture.