DETECÇÃO DE FALHAS EM LINHAS DE PLANTIO EM IMAGENS OBTIDAS POR VANT UTILIZANDO CNN E OPERADORES MORFOLÓGICOS

The world population grows every year, however, the arable lands of the planet are practically all in use or protected by environmental laws. Humanity needs to find ways to increase productivity in the countryside, and one of the ways is by making use of technology. This paper uses computational resources to detect failures in planting lines, through the analysis of plantation images obtained by UAVs. In the developed methodology, CNN, morphological operators and an algorithm were used to draw the planting lines. With the detected failures, the aim is to help rural producers to make better decisions, increase their production and reduce losses. The results obtained were satisfactory, but are closely linked to the quality of the image classification by CNN, which presented an F1 Score around 92%.