Análise semi-automática do arranjo espacial de plantas de milho utilizando visão computacional

• Ano de defesa: 2015
• Autor(a) principal: Brilhador, Anderson
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Tecnológica Federal do Paraná; Cornelio Procopio; Brasil; Programa de Pós-Graduação em Informática; UTFPR
• Palavras-chave em Português: Processamento eletrônico de dados; Agricultura de precisão; Biotecnologia agrícola; Electronic data processing; Precision farming; Agricultural biotechnology; CNPQ::CIENCIAS EXATAS E DA TERRA; Ciência da Computação
• Link de acesso: http://repositorio.utfpr.edu.br/jspui/handle/1/2954

Global demand for food is growing every year, requiring the development of new technologies that increase grain production without increasing the areas destined for planting. The corn crop is a major commodity in the world and is used as food, feed for other animals, in addition to having other industrial purposes. Corn is sensitive to the spatial arrangement of plants and any variation in distribution pattern can lead to reduction in the production of corn. Currently, the process of checking the uniformity of spacing between plants is done manually by agronomists and producers in order to predict possible production losses. In this context, this paper proposes an automatic approach to the analysis of the spatial arrangement of plants by measuring the spacing between corn plants in early stages of growth. From this measurement are extracted relevant information such as population density, uniformity of planting and loss estimates. The proposed approach uses computer vision techniques of low computational cost to identify corn plants and measure the spacing between plants, allowing its use in devices with low computational power such as smartphones and tablets. A set of images was built as an additional contribution of work, containing 222 panoramic images of corn planting in three conditions of planting: direct, conventional and direct after applying herbicides. The experimental results achieve 90% of rate accuracy and 87% sensitivity in identification of corn plants present on the base. A comparison of the measurements of the distances between plants made of manual and computer vision way, no presented significant differences between the measurements, indicating the effectiveness of the proposed approach at work.