Regionalização de área agrícola usando dados de imagens aéreas e coletas de campo

• Ano de defesa: 2020
• Autor(a) principal: Lorbieski, Rodrigo
• Orientador(a): Guedes, Luciana Pagliosa Carvalho
• Banca de defesa: Guedes, Luciana Pagliosa Carvalho , Dalposso, Gustavo Henrique , Uribe Opazo, Miguel Angel
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual do Oeste do Paraná; Cascavel
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Agrícola
• Departamento: Centro de Ciências Exatas e Tecnológicas
• País: Brasil
• Palavras-chave em Português: Agrupamentos; Árvore de decisão; Classificação supervisionada; Função núcleo-estimador
• Palavras-chave em Inglês: Groupings; Decision tree; Supervised classification; Core-estimator function
• Área do conhecimento CNPq: CIENCIAS EXATAS E DA TERRA
• Link de acesso: http://tede.unioeste.br/handle/tede/4866

Precision agriculture is an important tool that aims at optimizing the agricultural yield sector, however, its cost of implementation can be an obstacle for small farmers. An alternative would be regionalizing the agricultural area, and dividing it into plots or management zones, which can be worked individually according to the area characteristics. Multivariate data are common in designing these zones, thus, multivariate techniques for classification and grouping can be applied to these data, aiming at also taking into account their spatial information. Thus, this research outlined management areas for an agricultural area in four consecutive cropping years, using non-parametric hierarchical groupings, and considering spatial data information from physical and chemical attributes of soil, vegetative indexes and data of yield. Techniques were applied to reduce dimensionality and groupings of data, as well as geostatistical analyses and descriptive statistics. So, decision trees were built to better understand physical-chemical variables behavior in the different management zones formed, whose management zones are the response variable. The subsets that best formed the management zones varied from one cropping year to the next one, and their location was similar in three of the four cropping years analyzed. The excellent number of management zones was equal to two in all the studied cropping years. Decision trees proved to be important to characterize physical-chemical variables, as they helped to describe their distribution in the formation of each management zone.