Fenotipagem de alta eficiência no melhoramento genético da batata-doce por análise computacional de imagens

• Ano de defesa: 2022
• Autor(a) principal: Ana Clara Gonçalves Fernandes
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; Brasil; ICA - INSTITUTO DE CIÊNCIAS AGRÁRIAS; Programa de Pós-Graduação em Produção Vegetal; UFMG
• Palavras-chave em Português: Ipomoea batatas (L.) Lam..; Inteligência Computacional.; Melhoramento genético de hortaliças.; Análise de imagens.; Fenotipagem de alto rendimento.; Batata-doce; Inteligência Computacional; Processamento de imagens; Genética vegetal; Hortaliças > Melhoramento genético
• Link de acesso: http://hdl.handle.net/1843/49258; https://orcid.org/0000-0002-8161-8130

Sweet potato (Ipomoea batatas (L.) Lam..) stands out among the most planted vegetables in Brazil. Efforts for the genetic improvement of the crop are necessary, aiming to increase the productivity and quality of the roots. However, in selecting the best genotypes, it is necessary to evaluate a large number of characteristics, being an expensive and subjective process, making it difficult for the breeder to analyze it. In this sense, the adoption of new technologies to the phenotyping process represents a breakthrough and, among the possibilities, there is the image analysis associated with computational intelligence. Thus, it enables the evaluation of characteristics of interest, in a shorter period of time, reducing labor in the genetic improvement of the crop. In addition, it enables the classification of qualitative characters, reducing the subjectivity that exists in the grades given by the evaluators. Given the above, the objective is to apply methodologies for the computational automation of the evaluation of production and quality of sweet potato roots, through the analysis of digital images. They were evaluated sixteen half-sib progenies of sweet potato in a randomized block design, with four replications and ten plants per plot. For image acquisition, it was removed excess soil from the roots, and the images were acquired by digital camera and pre-processed in R software. They were evaluated at the plant level the following production and quality characteristics: individual root weight, root shape, damage caused by insects, coloring of peel and pulp. The use of image analysis associated with computer vision, as well as deep learning, is an efficient tool in sweet potato genetic improvement programs, helping in the phenotyping of the crop, as well as in decision making. In addition, they can be used in the development of applications and devices that help the producer in the classification of sweet potato roots. This makes it possible to use the vegetable for its most diverse purposes, reducing losses and waste in the production process of the crop.