Use este identificador para citar ou linkar para este item: http://repositorio.ufla.br/jspui/handle/1/58531
Título: Avaliação do efeito da altitude de voo e iluminação na classificação do uso do solo usando algoritmos de machine learning e imagens multiespectrais de Vant
Título(s) alternativo(s): Avaluation of the effect of flight altitude and lighting on land use classification using machine learning algorithms and multispectral uav images
Autores: Carvalho, Luís Marcelo Tavares de
Carvalho, Luís Marcelo Tavares de
Rocha, Samuel José Silva Soares da
Terra, Marcela de Castro Nunes Santos
Palavras-chave: Veículo aéreo não tripulado (VANT)
Floresta aleatória
Unmanned aerial vehicle (UAV)
Random forest
Geographic object-based image analysis (GEOBIA)
GEOBIA method
Support vector machine
Data do documento: 8-Nov-2023
Editor: Universidade Federal de Lavras
Citação: GARCIA, R. A. Avaliação do efeito da altitude de voo e iluminação na classificação do uso do solo usando algoritmos de machine learning e imagens multiespectrais de Vant. 2023. 72 p. Dissertação (Mestrado em Engenharia Florestal)–Universidade Federal de Lavras, Lavras, 2023.
Resumo: Information about land use and land cover in urban areas plays an important role in urban planning, providing essential insights for land suitability analysis, environmental assessments, and urban regeneration projects. Compared to satellite-based remote sensing, Unmanned Aerial Vehicle (UAV) remote sensing offers higher spatial and temporal resolution, making it a more effective method for land use classification. In this study, the influence of flight altitude (120 m and 150 m) and illumination (diffuse and direct) on land use classification was evaluated by combining an Object-Based Image Analysis (GEOBIA) approach with a machine learning algorithm using multispectral UAV images. Firstly, the images were segmented using the Multiresolution segmentation and Spectral Difference segmentation algorithms. Then, spectral, index, texture, and geometric features were combined to form schemes S1-S8. Finally, area classification was performed based on the eight schemes using the Random Forest (RF) and Support Vector Machine (SVM) classifiers. The results showed that the Random Forest classifier outperformed the Support Vector Machine in all schemes. Geometric features had a negative impact on the SVM classification accuracy, while the other three types of features had a positive impact. However, this behavior was not observed in RF, as the Random Forest classifier achieved an overall accuracy (AO) of 82% when combining spectral, texture, and geometric features (S6) for the image obtained at 150 m altitude under diffuse illumination.
Descrição: Arquivo retido, a pedido da autora, até novembro de 2024.
URI: http://repositorio.ufla.br/jspui/handle/1/58531
Aparece nas coleções:Engenharia Florestal - Mestrado (Dissertações)

Arquivos associados a este item:
Não existem arquivos associados a este item.


Este item está licenciada sob uma Licença Creative Commons Creative Commons