Método de fitopatometria utilizando processamento de imagens e Inteligência Artificial para detecção e quantificação de sintomas e sinais de Ferrugem Asiática da Soja aplicado ao melhoramento genético

• Ano de defesa: 2022
• Autor(a) principal: Silva, Diego Alves da
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Genética e Bioquímica
• Palavras-chave em Português: Ferrugem Asiática da Soja; Asian Soybean Rust; Fenotipagem por Imagem; Image Phenotyping; Fitopatometria; Phytopathometry; Processamento de Imagem; Image Processing; Melhoramento Genético; Genetic Breeding; CNPQ::CIENCIAS BIOLOGICAS::GENETICA::GENETICA VEGETAL; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO; CNPQ::CIENCIAS BIOLOGICAS::BIOQUIMICA; CNPQ::CIENCIAS AGRARIAS::AGRONOMIA::FITOSSANIDADE::FITOPATOLOGIA; Genética; Ferrugem asiática; Soja - Melhoramento genético; Processamento de imagens
• Link de acesso: https://repositorio.ufu.br/handle/123456789/36638; http://doi.org/10.14393/ufu.te.2022.5043

Soybeans are one of the most important crops in the world, with Brazil being one of the main players with its grain production and exports. Due to the increasing global demand for soybeans, a continuous increase in production is required, which could be affected by diseases such as Asian Soybean Rust (ASR) caused by the fungus Phakopsora pachyrhizi. Because of the diversity of races of the fungus and its several virulence causing genes, the process of genetic improvement focusing on resistance to ASR must be constant and applied in conjunction with chemical control and management techniques. The process of genetic breeding uses the severity of infection symptoms to identify resistant genotypes and is traditionally performed by the geneticist in the field in a visual manner. In this study, we propose a method for detecting and quantifying ASR symptoms using images that can be applied to the selection of genotypes with resistance levels. To develop the method, the work begins with a systematic review of the subject of study Artificial Intelligence and Image Processing, using the collected data to propose an exploratory study to understand the detection of image rust symptoms and develop an algorithm. The algorithm was validated against other existing methods, showing a correlation greater than 0.9, and software was developed to perform the quantification. An experiment was proposed to develop the method, focusing on understanding the effects of fungicides, the form of leaf collecting, and the form of image acquisition. The method and software were applied to the selection of genotypes from a population. In addition, another step of the work proposed the use of a convolutional neural network for classifying multiple pathogens in leaf areas.