Estimativa de produção do morangueiro por algoritmos de aprendizado de máquina a partir de imagens multiespectrais

• Ano de defesa: 2023
• Autor(a) principal: Oliveira, Larissa Silva de
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Uberlândia; Brasil; Programa de Pós-graduação em Agronomia
• Palavras-chave em Português: sensoriamento remoto; linear regression; support vector machine; remote sensing; CNPQ::CIENCIAS AGRARIAS::AGRONOMIA; Agronomia; Morango - Cultivo; Frutas - Cultivo; Aprendizado do computador; Imagens multiespectrais
• Link de acesso: https://repositorio.ufu.br/handle/123456789/37538; http://doi.org/10.14393/ufu.di.2023.7002

Estimating strawberry productivity is a manual, hard-working and subjective process. An efficient and accurate estimation process would allow better crop management. Thus, the objective of this work was to evaluate the performance of regression algorithms: Linear Regression and Support Vector Machine in estimating the number of fruits, average fruit mass and number of leaves of the strawberry plant through multispectral images obtained by remotely piloted aircraft (RPA). The experiment was carried out in the experimental area of the Botany Laboratory, at the Federal University of Uberlândia – Campus Monte Carmelo. The experimental design was randomized blocks with six treatments and four replications. The treatments consisted of six commercial strawberry cultivars: San Andreas, Albion, PR, Festival, Oso Grande and Guarani. The use of different cultivars aimed to generate genetic variability to test the estimation efficiency of the algorithms. Flights for image acquisition were performed weekly. The images passed through pre-processing, for the transmission of the radiometric values of each plant in the experimental area. These values were then used for training the production prediction algorithms. In the same period data were collected on average fruit mass, number of fruits per plant and number of leaves. The exciting results that the Linear Regression and Support Vector Machine algorithms are able to estimate in strawberry plants, through multispectral images transmitted by RPA, number of fruits (with accuracy of 99.55% and 84.26%, respectively), average mass of fruits (with accuracy of 99.91% and 89.62%, respectively), and number of leaves (with accuracy of 99.94% and 98.12%, respectively.