Detecção de Linhas e Falhas de Plantio por meio da Associação de um Algoritmo Genético para Multilimiarização à Transformada Discreta de Wavelet e Transformada de Hough Probabilística e como Mobile Cloud Computing pode Auxiliar na Melhoria de Desempenho

• Ano de defesa: 2020
• Autor(a) principal: Oliveira, Pedro Ventura 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 Engenharia Elétrica
• Palavras-chave em Português: Algoritmos Genéticos; Limiarização; Detecção de Linhas de Plantio; Transformada Discreta de Wavelet; Transformada de Hough; Visão Computacional; Mobile Cloud Computing; Genetic Algorithm; Thresholding; Crop Rows Detection; Discrete Wavelet Transform; Computer Vision; Hough Transform; Engenharia elétrica; CNPQ::ENGENHARIAS::ENGENHARIA ELETRICA; Algorítmos genéticos; Agricultura de precisão
• Link de acesso: https://repositorio.ufu.br/handle/123456789/30148; http://doi.org/10.14393/ufu.di.2020.685

In the field of Precision Agriculture, an issue that has been addressed to it are ways of detecting crop rows in georeferenced images, as well as failure in these rows on the plantation. The present work proposes the detection of these lines and flaws in a sugarcane plantation by applying a probabilistic Hough Transform performed after the pre-processing of sections of this plantation’s orthomosaic through a segmentation method based on the use of a Genetic Algorithm for multilimiarization , from the prior auxiliary application of a Discrete Wavelet Transform to the histograms of these sections. Furthermore, it is explored, conceptually, how the integration of the proposal to the Mobile Cloud Computing paradigm can improve the method in efficiency and accessibility of the data produced. With this, it is proposed the future integration of Mobile Cloud Computing, as a way to advance in efficiency of the method, in accessibility of the data obtained from the plantation and improvements that make the application in Precision Agriculture an even more profitable practice for the farmer. Previous studies have shown that experiments using different images and thresholding approaches, by means of comparative tests, are efficient in speed and accuracy on the results, which corroborates with the proposed computational method, and its application for the definition of the characteristics of interest in plantations is promising. The proposed computational algorithm demonstrated efficiency in its performance, when compared with other similar approaches, and is capable of detecting with good precision the rows and flaws in the sugarcane plantation used for the experimental tests. In this way, a future application of the Mobile Cloud Computing paradigm would be able to bring even more benefits in its overall efficiency, due to the robustness of the computational system that can be adopted.