GRUPOS DE MATURAÇÃO DA CULTURA DA SOJA AVALIADOS POR IMAGENS AÉREAS OBTIDAS POR VANT

Abstract

Soybeans are the most widely known oilseeds in the world, Brazil is important in its production. Precocious soybean cultivars are desirable, enabling cost reduction and a second crop. Genetic improvement companies are always creating better cultivars, using unmanned aerial vehicles for areas mapped by aerial images without direct contact with the plant which help the area evaluation. The objective was to evaluate the use of aerial images to assess plant maturation. Eight flights were carried out in a plot of the Retiro farm in the municipality of Rio Verde-GO, with the Phantom 4 drone coupled with an RGB sensor. The images were processed in the Laboratory of Drones and Vant in Agriculture of the IF Goiano, Rio Verde campus, using the Pix4D software and analyzed in the QGIS software. The evaluated agronomic character was the maturation group, the data were also obtained in the field with visual assessment of the plants following the comparison of control plants with known maturation group. A boxplot analysis was performed, then a correlation was made between the RGB reflectance and the PRI index with Maturation Group (GM), simple and multiple linear regression analysis. The data were randomly divided into two groups being one with 80% of the genotypes and the other with 20%. To obtain the models that were performed in the first group, a multiple linear regression with all wavelengths and indices and a stepwise multiple linear regression with variable selection model were carried out. The variable selection criterion used was the AIC. The regression stepwise shows the highest R2 among the proposed models, which was 39.94%. It was concluded that the stepwise model was the best method for evaluating soybean maturation groups, but it is not sufficient for greater data reliability.