Redes neurais em processamento de imagens para detecção de mariposas da broca da cana-de-açúcar em armadilhas

Neural networks for image processing is a field of study with increasing applications for society in general, as well as industries and agribusiness. There are several challenges in agribusiness whose use of image processing techniques can contribute to the partial or complete solution of problems, one of these challenges being pest control. The main objective of this work is to develop a neural network model capable of detecting moths present in traps installed in sugar cane farms. A dataset with images of moths in traps was used to train the network, after which the network performance data were consolidated to enable the analysis of its performance in detecting these objects. The developed solution achieved a detection percentile of 88%, in addition to mapping strategies to make the neural network even more robust and reliable, which shows the viability of using image processing techniques with neural networks for this type of application.