Técnicas de agricultura digital para predição da maturação do amendoim
Carregando...
Arquivos
Data
2021-07-29
Autores
Souza, Jarlyson Brunno Costa
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Técnicas como o Sensoriamento Remoto (SR) e Aprendizado de Máquinas estão sendo amplamente utilizadas no estudo da vegetação e têm mostrado potencial em seu uso para monitoramento e estimativa de diversos parâmetros agronômicos. No entanto, o uso destas técnicas para a cultura do amendoim ainda se encontra incipiente. Dessa forma, os objetivos deste trabalho foram: (i) avaliar a qualidade das plataformas de sensoriamento remoto por meio de cartas de controle e verificar qual tem melhor potencial para o monitoramento da cultura. (ii) predizer a maturação do amendoim utilizando técnicas de Redes Neurais Artificias (RNA) e SR. No primeiro capítulo encontra-se a revisão de literatura. No segundo, realizou-se análise de variabilidade com auxílio de cartas de controle para verificar a qualidade da reflectância espectral e dos índices de vegetação gerados a partir de imagens de Veículo Aéreo Não Tripulado (UAV) e Satélite de alta resolução. Os resultados demonstraram que o satélite apresenta melhor qualidade no monitoramento espaço-temporal para as bandas e índices de vegetação, enquanto que o UAV saturou todas a bandas ao longo do tempo. No terceiro capítulo foram utilizados modelos de RNA’S para predição da maturação do amendoim, utilizando imagens de UAV e satélite. Verificou-se potencial nas duas plataformas para predizer a maturação, com os índices de vegetação NDVI do satélite e NDRE do UAV. Os modelos de redes MLP e RBF obtiveram resultados semelhantes em relação ao desempenho dos modelos gerados. Por fim, no capítulo cinco, tem-se as considerações finais, em que são apresentadas recomendações e trabalhos futuros para contribuir para o avanço do conhecimento.
Techniques such as Remote Sensing (RS) and Machine Learning are being widely used in the study of vegetation and have shown potential in their use for monitoring and estimation of several agronomic parameters. However, the use of these techniques for peanut cultivation is still incipient. Thus, the objectives of this work were (i) evaluate the quality of remote sensing platforms by means of control charts and verify which has the best potential for crop monitoring. (ii) to predict peanut maturity using Artificial Neural Networks (ANN) and SR techniques. The first chapter contains the literature review. In the second, variability analysis was performed using control charts to verify the quality of spectral reflectance and vegetation indices generated from Unmanned Aerial Vehicle (UAV) and high-resolution satellite images. The results showed that the satellite has better spatio-temporal monitoring quality for the bands and vegetation indices, while the UAV saturated all bands over time. In the third chapter, ANN models were used for peanut maturity prediction, using UAV and satellite images. Potential was found in both platforms to predict maturity, with the satellite NDVI and UAV NDRE vegetation indices. The MLP and RBF network models obtained similar results regarding the performance of the generated models. Finally, in chapter five, there are the final considerations, in which recommendations and future work are presented to contribute to the advancement of knowledge.
Techniques such as Remote Sensing (RS) and Machine Learning are being widely used in the study of vegetation and have shown potential in their use for monitoring and estimation of several agronomic parameters. However, the use of these techniques for peanut cultivation is still incipient. Thus, the objectives of this work were (i) evaluate the quality of remote sensing platforms by means of control charts and verify which has the best potential for crop monitoring. (ii) to predict peanut maturity using Artificial Neural Networks (ANN) and SR techniques. The first chapter contains the literature review. In the second, variability analysis was performed using control charts to verify the quality of spectral reflectance and vegetation indices generated from Unmanned Aerial Vehicle (UAV) and high-resolution satellite images. The results showed that the satellite has better spatio-temporal monitoring quality for the bands and vegetation indices, while the UAV saturated all bands over time. In the third chapter, ANN models were used for peanut maturity prediction, using UAV and satellite images. Potential was found in both platforms to predict maturity, with the satellite NDVI and UAV NDRE vegetation indices. The MLP and RBF network models obtained similar results regarding the performance of the generated models. Finally, in chapter five, there are the final considerations, in which recommendations and future work are presented to contribute to the advancement of knowledge.
Descrição
Palavras-chave
Sensoriamento remoto, Aprendizado, Artificiais, Controle de qualidade