Visão computacional aplicada à avaliação da cor do tegumento em feijão carioca

• Ano de defesa: 2021
• Autor(a) principal: Silva, Lorena Caroline Dumbá
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Lavras; Programa de Pós-Graduação em Genética e Melhoramento de Plantas; UFLA; brasil; Departamento de Biologia
• Palavras-chave em Português: Feijoeiro - Melhoramento genético; Feijão - Melhoramento genético; Sistemas computacionais; Imagens digitais - Análise; Common Bean - Breeding; Computer systems; Digital images - Analysis; Melhoramento Vegetal
• Link de acesso: http://repositorio.ufla.br/jspui/handle/1/47986

The common bean (Phaseolus vulgaris L.) is one of the crops that present great variability of grains in terms of color, size, and shape. These particularities are highly valued by the consumer and the industry, being fundamental for the acceptance of a new cultivar. The carioca beans are the most cultivated and consumed in Brazil, with more than 70% of the national production, which is why the improvement programs in Brazil are dedicated, especially, to the development of cultivars from this commercial group. The producers, cereal producers, as well as consumers are looking for a carioca bean with a lighter color and with late darkening of the grains, as they associate the darkening of the tegument with old grains that are difficult to cook, resulting in a low-quality product with lower value market. The main ways of evaluating the color of beans are visually or through equipment such as a colorimeter. The evaluations are carried out using a grading scale, therefore, they do not present the same accuracy when compared to those obtained in an equipment. With the colorimeter, however, we can have limitations in the quality of specifications, in the quantity and variety of data that can be sought. New approaches, such as computer vision, offer solutions to increase the efficiency and effectiveness of the improvement for this character. This approach allows to automatically measure a variety of high-resolution images. Digital images are visual information stored in a digital form, which can be viewed in computer systems. As this type of image is stored in a binary form in computer systems, they are not available for environmental impact, which could lead to loss of quality. The digital processing of these images, on the other hand, is an area of science that seeks to analyze, process, or improve digital images to obtain information about objects of interest.