@MastersThesis{Arcoverde:2008:EsEsEs,
author = "Arcoverde, Gustavo Felipe Balu{\'e}",
title = "Estratifica{\c{c}}{\~a}o espacial para estimativa da {\'a}rea
de culturas agr{\'{\i}}colas de ver{\~a}o com imagens de
pr{\'e}-plantio",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2008",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2008-08-27",
keywords = "Estimativa de {\'a}rea agr{\'{\i}}cola,
estratifica{\c{c}}{\~a}o espacial, reconhecimento de
padr{\~o}es em imagens de sat{\'e}lite, mapas
auto-organiz{\'a}veis, {\'a}rvore de decis{\~a}o, agricultural
crop area estimation, spatial stratification, pattern recognition
iin satellite images, self-organizing maps, tree decision.",
abstract = "O levantamento da {\'a}rea agr{\'{\i}}cola {\'e} uma
informa{\c{c}}{\~a}o fundamental para o planejamento
agr{\'{\i}}cola, econ{\^o}mico, agr{\'a}rio, ambiental, ou
social. Recomenda-se que o c{\'a}lculo da {\'a}rea
agr{\'{\i}}cola seja por vias probabil{\'{\i}}sticas e de
forma estratificada. A estratifica{\c{c}}{\~a}o tem como
principais benef{\'{\i}}cios a redu{\c{c}}{\~a}o da
vari{\^a}ncia global e a otimiza{\c{c}}{\~a}o da
verifica{\c{c}}{\~a}o da verdade amostral. A
estratifica{\c{c}}{\~a}o espacial comumente {\'e} realizada por
interpreta{\c{c}}{\~a}o visual de dados cartogr{\'a}ficos e
imagens de sat{\'e}lite, visando identificar padr{\~o}es de uso
da terra com probabilidades de ocorr{\^e}ncia de determinada
cultura agr{\'{\i}}cola. Este trabalho teve como objetivo
avaliar estratifica{\c{c}}{\~o}es semi-autom{\'a}ticas para
estimativas de {\'a}reas com uso agr{\'{\i}}cola de culturas de
ver{\~a}o no munic{\'{\i}}pio de Barretos (SP). Para tanto
utilizaram-se duas imagens do sensor TM/Landsat-5 de datas de
pr{\'e}-plantio de culturas agr{\'{\i}}colas de ver{\~a}o. Os
estratos foram gerados pela identifica{\c{c}}{\~a}o e
classifica{\c{c}}{\~a}o de padr{\~o}es de cobertura da terra
nas datas de pr{\'e}-plantio, por an{\'a}lise espectral por
pixels (AEP), e por an{\'a}lise de multi-atributos por objeto
(AMO). Na AEP empregaram-se modelo linear de mistura espectral,
an{\'a}lise de vetor de mudan{\c{c}}as, e an{\'a}lise de Mapas
de Regras por m{\'a}xima verossimilhan{\c{c}}a. A AMO foi
abastecida por diferentes atributos extra{\'{\i}}dos pelo
programa DEFINIENS, e empregou classificadores por mapas
auto-organiz{\'a}veis (SOM), {\'a}rvore de decis{\~a}o por
algoritmo C 4.5, e com associa{\c{c}}{\~a}o de ambos. As
estratifica{\c{c}}{\~o}es geradas foram consideradas em sua
maioria como regulares. A principal causa deste desempenho regular
foi a confus{\~a}o que os classificadores apresentaram em
rela{\c{c}}{\~a}o {\`a}s {\'a}reas de
cana-de-a{\c{c}}{\'u}car planta. Houve redu{\c{c}}{\~a}o da
vari{\^a}ncia global da estimativa de {\'a}rea que utilizou os
estratos com maior {\'{\i}}ndice kappa, em
compara{\c{c}}{\~a}o a uma estimativa de {\'a}rea por
amostragem aleat{\'o}ria simples. No entanto, nesta mesma
compara{\c{c}}{\~a}o, n{\~a}o haveria otimiza{\c{c}}{\~a}o
para a verifica{\c{c}}{\~a}o da verdade amostral, posto que foi
necess{\'a}rio atribuir mais elementos amostrais fora dos
estratos definidos para a redu{\c{c}}{\~a}o da vari{\^a}ncia
global. Concluiu-se que estratifica{\c{c}}{\~o}es espaciais
realizadas pelas abordagens adotadas, realizadas em regi{\~o}es
cuja {\'a}rea de interesse se apresente de forma fragmentada e/ou
com pouca representa{\c{c}}{\~a}o na {\'a}rea total, n{\~a}o
s{\~a}o indicadas para a redu{\c{c}}{\~a}o da vari{\^a}ncia
global e para a otimiza{\c{c}}{\~a}o da verifica{\c{c}}{\~a}o
da verdade amostral de forma associada. ABSTRACT: Agricultural
crop surveys are fundamental information for planning of
agriculture, economics, agraro-environment, or social. It is
recommended that the calculation of the agriculture crop area be
made by probabilistic methods and using stratification procedures.
The stratification has as main benefits the decrease of total
variance and the quality improvement of sampling check. The
spatial stratification is usually carried out by visual
interpretation of cartographic data and satellite images, aiming
to identify patterns of land use with probabilities of occurrence
of specific agriculture crop. This study has the objective to
evaluate semi-automatic stratifications for area estimates of
summer agricultural crops in the municipality of Barretos, State
of S{\~a}o Paulo, Brazil. For this purpose, two TM/Landsat-5
images from pre-planting dates of summer agricultural crops were
used. The strata were generated by the identification and
classification of land cover patterns in pre-planting dates, by
pixel spectral analysis (PSA) and by multi-attributes object
analysis (MOA). In PSA unmixing linear modeling, change vector
analysis and Rule Maps analysis by maximum likelihood were used.
The MOA was supplied with different attributes extracted from
DEFINIENS software, and employed classifiers by self-organizing
maps (SOM), decision tree by C 4.5 algorithm as well as the
association of both. Most of the stratifications generated were
considered to be regular. The main cause of this regular
performance was due to the confusion presented by the classifiers
in relation to the new planted sugar cane areas. There was a
decrease in the total variance of the area estimates that used the
strata with higher kappa index, in comparison to an area estimate
by simple random sampling. However, there wouldnt be improvement
in checking the sampling quality because it was necessary to
allocate more sampling elements outside the defined strata in
order to decrease the total variance estimate. It was concluded
that spatial stratifications adopted here, which were carried out
in regions where interest areas are presented as fragmented spots
and/or with little proportions of the total area, are not
indicated for the reduction of the total variance and for the
improvement of checking the sampling quality.",
committee = "Renn{\'o}, Camilo Daleles (presidente) and Epiphanio, Jos{\'e}
Carlos Neves (orientador) and Formaggio, Ant{\^o}nio Roberto and
Fonseca, Leila Maria Garcia and Rocha, Jansle Vieira",
copyholder = "SID/SCD",
englishtitle = "Spatial stratification for area estimate of summer agricultural
crops using pre-planting images",
language = "pt",
pages = "180",
ibi = "8JMKD3MGP8W/3448EQH",
url = "http://urlib.net/ibi/8JMKD3MGP8W/3448EQH",
targetfile = "paginadeacesso.htm",
urlaccessdate = "02 maio 2024"
}