Estimation of failures in soybean crops from aerial images obtained by RPA

The dissemination of Remotely Piloted Aircraft (RPA) in the agricultural sector made it possible to map crop failures and disease incidence. This work aims to estimate crop failures. For that purpose, we conducted three experiments with a soybean crop at the IFSULDEMINAS school farm - Inconfidentes Campus, by carrying out an aerial survey with an RPA, which generated an orthophoto of the area of interest. To quantify the failures existing in the experiments, we carried out a supervised classification to distinguish the soybean plants of the exposed soil. After the classification, the kappa index was calculated to verify whether the classification was satisfactory. With this, it was possible to calculate the percentage of failures obtained in each plot of the experiment. Finally, we analyzed the variance to verify if the percentage of failures of each plot had significant differences between them. We observed that in two experiments, there was a statistical difference in the number of failures, and in one experiment there was no difference.