Uso de imagens aéreas com drones para identificação de falhas no estabelecimento da soja

Abstract

Precision agriculture uses advanced technologies that, through data collection and analysis, allow the farmer to make the best decisionmaking management, with the aim of greater production and cost reduction. The use of UAV - Unmanned Aerial Vehicle in agriculture has become more and more common, and can bring many benefits to the agricultural industry. Allows the monitoring and analysis of crops, with high precision and speed. It makes it possible to monitor the development of crops, identify pests and diseases, collecting data through sensors and camera images. These data and images, when analyzed by specific software, allow the identification of areas of a plot that present a problem in the establishment of the crop. The objective was to evaluate the area of an agricultural plot through images, in the visible spectrum (RGB), taken with a drone, generating vegetation indexes and classification of images in the identification of failures in the establishment of soybeans. Images acquired by drones were used, and through geoprocessing techniques, an orthomosaic was created. This allowed the integration of these images in the QGIS software, in which several indices such as GLI, NGDRI and RI, which are related to vegetation, were calculated. Subsequently, the images were submitted to supervised classification and binary thresholding processes. This approach enabled the calculation of the Kappa coefficient, which was used to assess the accuracy of the method and to identify areas in the stands where plant crushing and faults occurred. Combining GLI and RI indices provides more consistent and accurate analysis compared to combining GLI and NGRDI, as reflected by Kappa coefficients and percentages of hits. Considering the Binary Threshold methods, it can be observed that they were efficient, mainly maximum/minimum or three times the standard deviation when compared to the Random Forest algorithm. Thus, being able to clearly identify the failures arising from the development of the crop and traces of crushing arising from the application of pesticides in the study plot.