@PhDThesis{Mello:2013:SpBaMe,
author = "Mello, Marcio Pupin",
title = "Spectral-temporal and Bayesian methods for agricultural remote
sensing data analysis",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2013",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2013-08-19",
keywords = "spectral-temporal response surface, sugarcane pre-harvest burning,
Bayesian Network, plausible reasoning, soybean mapping,
superf{\'{\i}}cie de resposta espectro-temporal, queima da palha
na pr{\'e}-colheita da cana-de-a{\c{c}}{\'u}car, rede
Bayesiana, l{\'o}gica racional, mapeamento da soja.",
abstract = "Informa{\c{c}}{\~o}es agr{\'{\i}}colas confi{\'a}veis tem se
tornado cada vez mais importantes para os tomadores de
decis{\~o}es. Especialmente quando s{\~a}o obtidas em tempo
h{\'a}bil, essas informa{\c{c}}{\~o}es s{\~a}o altamente
relevantes para o planejamento estrat{\'e}gico do pa{\'{\i}}s.
Apesar de o sensoriamento remoto mostrar-se promissor para
aplica{\c{c}}{\~o}es em mapeamento agr{\'{\i}}cola, com
potencial de melhorar as estat{\'{\i}}sticas agr{\'{\i}}colas
oficiais, esse potencial n{\~a}o tem sido amplamente explorado.
Existem poucos exemplos bem sucedidos do uso operacional do
sensoriamento remoto para mapeamento sistem{\'a}tico de culturas
agr{\'{\i}}colas e, para garantir resultados precisos, eles
s{\~a}o fortemente baseados em interpreta{\c{c}}{\~a}o visual
de imagens. De fato, apesar dos substanciais avan{\c{c}}os em
an{\'a}lise de dados de sensoriamento remoto, novas t{\'e}cnicas
para automatizar a an{\'a}lise de dados em sensoriamento remoto
com aplica{\c{c}}{\~o}es agr{\'{\i}}colas s{\~a}o
desej{\'a}veis, especialmente no prop{\'o}sito de manter a
consist{\^e}ncia e a precis{\~a}o dos resultados. Neste
contexto, existe uma demanda crescente pelo desenvolvimento e
implementa{\c{c}}{\~a}o de m{\'e}todos automatizados de
an{\'a}lise de dados de sensoriamento remoto com
aplica{\c{c}}{\~o}es em agricultura. Assim, o principal objetivo
desta tese {\'e} propor o desenvolvimento e a
implementa{\c{c}}{\~a}o de m{\'e}todos para automatizar a
an{\'a}lise de dados de sensoriamento remoto em
aplica{\c{c}}{\~o}es agr{\'{\i}}colas, com foco na
consist{\^e}ncia e precis{\~a}o dos resultados. Este documento
foi escrito como uma cole{\c{c}}{\~a}o de dois artigos, cada um
com foco nos seguintes pontos: (i) an{\'a}lise multitemporal,
multiespectral e multisensor, permitindo a descri{\c{c}}{\~a}o
das varia{\c{c}}{\~o}es espectrais de alvos agr{\'{\i}}colas
ao longo do tempo; e (ii) intelig{\^e}ncia artificial na
modelagem de fen{\^o}menos usando dados de sensoriamento remoto e
informa{\c{c}}{\~o}es complementares de maneira integrada. Dois
estudos de caso referentes ao mapeamento da colheita da cana em
S{\~a}o Paulo e ao mapeamento da soja no Mato Grosso foram usados
para testar as metodologias batizadas de STARS e BayNeRD,
respectivamente. Os resultados dos testes confirmaram que ambos os
m{\'e}todos propostos foram capazes de automatizar processos de
an{\'a}lises de dados de sensoriamento remoto com
aplica{\c{c}}{\~o}es agr{\'{\i}}colas, com consist{\^e}ncia e
precis{\~a}o. ABSTRACT: Reliable agricultural statistics has
become increasingly important to decision makers. Especially when
timely obtained, agricultural information is highly relevant to
the strategic planning of the country. Although remote sensing
shows to be of great potential for agricultural mapping
applications, with the benefit of further improving official
agricultural statistics, its potential has not been fully
explored. There are very few successful examples of operational
remote sensing application for systematic mapping of agricultural
crops, and they are strongly supported by visual image
interpretation to allow accurate results. Indeed, despite the
substantial advances in remote sensing data analysis, techniques
to automate remote sensing data analysis focusing on agricultural
mapping applications are highly valuable but have to maintain
consistency and accuracy. In this context, there continues to be a
demand for development and implementation of computer aided
methods to automate the processes of analyzing remote sensing
datasets for agriculture applications. Thus, the main objective of
this thesis is to propose implementation of computer aided
methodologies to automate, maintaining consistency and accuracy,
processes of remote sensing data analyses focused on agricultural
thematic mapping applications. This thesis was written as a
collection of two papers related to a core theme, each addressing
the following main points: (i) multitemporal, multispectral and
multisensor image analysis that allow the description of spectral
changes of agricultural targets over time; and (ii) artificial
intelligence in modeling phenomena using remote sensing and
ancillary data. Study cases of sugarcane harvest in S{\~a}o Paulo
and soybean mapping in Mato Grosso were used to test the proposed
methods named STARS and BayNeRD, respectively. The two methods
developed and tested confirm that remotely sensed (and ancillary)
data analysis can be automated with computer aided methods to
model a range of cropland phenomena for agriculture applications,
maintaining consistency and accuracy.",
committee = "Formaggio, Antonio Roberto (presidente) and Rudorff, Bernado
Friedrich Theodor (orientador) and Santos, Rafael Duarte Coelho
dos and Batista, Get{\'u}lio Teixeira and Vieira, Carlos
Ant{\^o}nio Oliveira",
copyholder = "SID/SCD",
englishtitle = "M{\'e}todos Espectro-temporal e Bayesiano para an{\'a}lise de
dados em sensoriamento remoto agr{\'{\i}}cola",
language = "en",
pages = "120",
ibi = "8JMKD3MGP7W/3ERM89S",
url = "http://urlib.net/ibi/8JMKD3MGP7W/3ERM89S",
targetfile = "publicacao.pdf",
urlaccessdate = "09 maio 2024"
}