Estimativa da produtividade de milho (Zea Mays L.) através de imagens obtidas por veículo aéreo não tripulado

Abstract

Get the determination of grain yield and its variability is important in agriculture because it enables the understanding of factors that limit this income, giving more benefits to decision-making in the management of the tillage. Precision agriculture uses the yield mapping to determine the yield and spatial variability in the field. Unmanned aerial vehicles (UAV's) embedded with multispectral sensors became a potential tool for monitoring and identification of the spatial variability of yield in the field. This study had as objective to evaluate the efficiency of use of multispectral images from UAV to estimate the yield of corn (Zea mays L.). The survey was conducted in a field of 51.6 ha, under no-tillage and precision agriculture, at Boa Vista farm, located in São Martinho da Serra, Rio Grande do Sul. The soil is Neosolo and climate Cfa, according to Köppen classification. The culture used in the experiment was corn hybrid Pioneer 1630 HX, with spacing of 0.5 meters and 70.000 plants per hectare, sowing in 08/20/2014 and harvesting in 20/01/2015. The field image capture ocurred with the UAV model EI Asesor / 5 equipped with two CMOS sensors: the first, multispectral model Teracam ACD Micro, with 3 spectral bands: a band of green, red and near infrared, and the second sensor, model Flir Tau 2, with the spectral band of thermal infrared. It was generated the orthorectified mosaic multispectral images with spatial resolution of 0.7 meters and in sequence the normalized vegetation index (NDVI). Yield data were obtained through the John Deere 9670 combine, boarded with the monitor kit and harvesting sensors brand and model Trimble FMX. To evaluate the correlation of yield data with the multispectral image and the NDVI obtained with UAV was used R² Pearson, where were sampled 200 points stratified into 4 yield classes, being 50 points per class. To evaluate the correlation of corn yield with the image and NDVI index was used Pearson R², together with a visual evaluation. Yield data crossed with: the green band resulted in the linear equation with R² = 0.05, with low correlation; the infrared band near resulted in the linear equation with R² = 0.36, with an average correlation; the band Red resulted in the exponential equation with R² = 0.38, with an average correlation; the band in the thermal infrared resulted in negative linear equation with R² = 0.68, with high correlation; the NDVI resulted in the linear equation with R² = 0.75, with high correlation, and visual analysis of NDVI with the yield map also showed consistent results with statistical analysis. Therefore, was found a significant linear regression between NDVI vegetation index and corn yield, being possible to estimate corn yield potential through the UAV images, which provide monitoring of corn yield beforehand the harvest, confirming its importance for the precision agriculture.