Desenvolvimento de um sistema de visão computacional para fenotipagem de alta precisão

• Ano de defesa: 2017
• Autor(a) principal: Santos, Marcos Roberto dos
• Orientador(a): Rieder, Rafael
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade de Passo Fundo
• Programa de Pós-Graduação: Programa de Pós-Graduação em Computação Aplicada
• Departamento: Instituto de Ciências Exatas e Geociências – ICEG
• País: BR
• Palavras-chave em Português: Aparelhos e materiais eletrônicos; Biomassa; Processamento de imagens; Visão por computador
• Palavras-chave em Inglês: Electronic apparatus and materials; Biomass; Image processing; Computer vision
• Área do conhecimento CNPq: CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: http://10.0.217.128:8080/jspui/handle/tede/28

The use of computational techniques to obtain and analyze information from the plant phenotype allow to increase the scientific interest and improve the data interpretation. With this in mind, techniques of computer vision and image processing can be used to get accuracy data regularly, for instance, vegetation indexes - and, at the same time, enable the correlation of this data with the biomass and the production of a plant species. In this context, this work presents the development a precision phenotyping platform using computer vision resources for a wheat crop, including three software: an image collector, concatenated to an Appliance that provides a controlled environment; an image processing application to extract new data; and a web solution for final view of results. As a case study, we followed and used data from an experiment under the responsibility of the Embrapa Trigo, in a farm named Capão Bonito, based in Carazinho, RS. Our regression analysis showed that NDVI variable explains 98,9, 92,8 e 88,2% of the variability founded on Biomass values for the treatments with 82, 150 e 200 kg de Nho􀀀1, respectively. Consequently, the NDVI obtained by sensors presented significant relation with the production for the three phenological stages, pointing to the possibility of elaboration of a productivity prediction model, which could be used since the beginning of planting.