• JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
  • JoomlaWorks Simple Image Rotator
 
  Bookmark and Share
 
 
Doctoral Thesis
DOI
https://doi.org/10.11606/T.11.2021.tde-26052021-144804
Document
Author
Full name
Leonardo Leite de Melo
E-mail
Institute/School/College
Knowledge Area
Date of Defense
Published
Piracicaba, 2021
Supervisor
Committee
Marques, Patricia Angélica Alves (President)
Barros, Timóteo Herculino da Silva
Frizzone, Jose Antonio
Romero, Roseli Aparecida Francelin
Title in Portuguese
Deep learning para identificação de déficit hídrico em plantas com base em imagens térmicas
Keywords in Portuguese
Imagem termal
Manejo de irrigação
Redes Neurais
Transfer learning
Abstract in Portuguese
O uso racional de recursos na agricultura ganhou importância nos últimos anos devido à necessidade de assegurar a sustentabilidade da produção agrícola, de forma a evitar consequências ambientais e até mesmo a iminência de escassez de recursos, como é o caso da água. Pela complexidade de estimar a resposta da planta à disponibilidade hídrica, uma técnica que vem obtendo grande importância é a utilização de imagens térmicas. Porém, o diagnóstico por imagens não é intuitivo e demanda do avaliador conhecimentos físico-químicos do solo e da planta, além de tempo e experiência. Devido a tal limitação, buscaram-se métodos computacionais que possam ser utilizados para realizar essa tarefa, estimando o estado hídrico de plantas a partir de imagens térmicas, suprindo a necessidade de um especialista. Para isso, duas técnicas de eficácia comprovada foram utilizadas: a rede Inception-Resnet-v2 e a técnica transfer learning. Experimentos foram realizados e os resultados obtidos mostram que o sistema de classificação do estresse hídrico na planta desenvolvido, com avaliação a partir da imagem térmica de modo não destrutivo, alcançou um desempenho superior em comparação à avaliação feita por especialista. Além disso, o desempenho foi superior na acurácia global, bem como em sua capacidade de distinguir entre as classes, demonstrando ser uma ferramenta eficaz para a realização de tal tarefa, demandando menor tempo.
Title in English
Deep Learning for identification of water deficit in plants based on thermal images
Keywords in English
Irrigation management
Neural network
Thermal image
Transfer learning
Abstract in English
Rationing of resources in agriculture has gained focus in recent years, owing to the need to ensure the sustainability of agricultural production, to avoid unfavorable environmental consequences with the imminent scarcity of resources such as water. The use of thermal images to evaluate water availability of plants has been gaining attention recently owing to the complexity of estimating a plant's response to water availability. However, diagnostic imaging is not intuitive and requires evaluator knowledge of the physicochemical properties of the soil and the plant species, in addition to time and experience. To circumvent this limitation, computational methods can be used to perform this task without the requirement of a specialist. The objective of this study was to develop a method to estimate the water availability of sugarcane plants using thermal images, without the need of a specialist. For this, two neural network methods with proven effectiveness were implemented: Inception-Resnet-v2 network and transfer learning technique. Experiments were conducted and the results showed that the developed system achieved superior performance compared to the assessments made by a specialist and aided in classifying the water stress of plant's thermal images in a nondestructive manner. In addition to this overall superior performance in accuracy, the neural network demonstrated a greater ability to distinguish between the classes of thermal stress. Thus, the system developed in this study is a less time-consuming, affordable, and effective tool for estimating the water availability.
 
WARNING - Viewing this document is conditioned on your acceptance of the following terms of use:
This document is only for private use for research and teaching activities. Reproduction for commercial use is forbidden. This rights cover the whole data about this document as well as its contents. Any uses or copies of this document in whole or in part must include the author's name.
Publishing Date
2021-05-27
 
WARNING: Learn what derived works are clicking here.
All rights of the thesis/dissertation are from the authors
CeTI-SC/STI
Digital Library of Theses and Dissertations of USP. Copyright © 2001-2024. All rights reserved.