DIAGNOFOMIL: FERRAMENTA PARA DIAGNÓSTICO DE DOENÇAS FÚNGICAS EM FOLHA DE MILHO

• Autor(a) principal: Stahnke, Eduardo
• Data de Publicação: 2013
• Outros Autores: Santos, Fernando dos
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Revista Brasileira de Contabilidade e Gestão
• Texto Completo: https://www.revistas.udesc.br/index.php/reavi/article/view/4083
• topic: Doença do Milho; Visão Computacional; Redes Neurais; Corn Diseases; Computer Vision; Neural Networks; Enfermedad del Maiz; Visión por Computador; Redes Neuronales

This paper presents the development of a tool for diagnosis of fungal diseases in corn leaves. The tool uses techniques of image processing and computer vision to extract the characteristics from images. These characteristics are extracted using threshold segmentation over the H plane of the image in the HSV color model, highlighting the disease spot. Given the disease spots, their Hu moments are calculated and used as inputs to a neural network. To make the diagnosis a multilayer perceptron neural network is used. Backpropagation is used for the network training. The system achieved an acceptable level of accuracy, diagnosing correctly 100% of the White Spot disease samples, 80% of the Cercospora samples and 80% of the Anthracnose samples.