@MastersThesis{Carvalho:2011:ClCoUs,
author = "Carvalho, Marcus Vin{\'{\i}}cius Alves de",
title = "Classifica{\c{c}}{\~a}o da cobertura e uso do solo com imagens
Worldview-2 utilizando minera{\c{c}}{\~a}o de dados e
an{\'a}lise baseada em objetos",
school = "Instituto Nacional de Pesquisas Espaciais",
year = "2011",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2011-09-26",
keywords = "sensoriamento remoto, uso e cobertura do solo urbano,
minera{\c{c}}{\~a}o de dados, an{\'a}lise baseada em objetos
geogr{\'a}ficos, imagens de alta resolu{\c{c}}{\~a}o espacial,
remote sensing, urban land use and land cover, data mining,
geographic object-based image analysis, high spatial resolution
images.",
abstract = "Os produtos provenientes dos novos sensores orbitais de alta
resolu{\c{c}}{\~a}o apresentam consider{\'a}vel potencial em
estudos e aplica{\c{c}}{\~o}es sobre o ambiente urbano. A
utiliza{\c{c}}{\~a}o dos dados destes sistemas sensores abre
novas perspectivas para a extra{\c{c}}{\~a}o autom{\'a}tica de
relevantes informa{\c{c}}{\~o}es para o planejamento e
gest{\~a}o do ambiente. No entanto, para a eficiente
aquisi{\c{c}}{\~a}o da informa{\c{c}}{\~a}o, s{\~a}o
necess{\'a}rias concep{\c{c}}{\~o}es inovadoras tanto na etapa
de segmenta{\c{c}}{\~a}o, quanto na distin{\c{c}}{\~a}o dos
objetos que comp{\~o}em a imagem. Neste estudo, utilizaram-se
imagens do sat{\'e}lite WorldView-2, uma nova
concep{\c{c}}{\~a}o em observa{\c{c}}{\~a}o da Terra porque
al{\'e}m da alta resolu{\c{c}}{\~a}o espacial (0,50 m),
apresenta oito bandas multiespectrais, tr{\^e}s bandas na faixa
do vis{\'{\i}}vel (Red, Green e Blue) e a banda do Infravermelho
Pr{\'o}ximo (NIR), al{\'e}m das quatro novas bandas: Coastal
(400-450 nm), Yellow (585-625 nm), Red Edge (705-745 nm) e
Near-Infrared-2 (860-1040 nm). Assim, o objetivo deste trabalho
foi avaliar os melhores atributos e limiares para o mapeamento da
cobertura e uso do solo em imagens WorldView-2. O experimento foi
realizado em um setor a oeste da Regi{\~a}o Metropolitana de
S{\~a}o Paulo. A classifica{\c{c}}{\~a}o da cobertura do solo
foi realizada em tr{\^e}s etapas: 1) gera{\c{c}}{\~a}o do
modelo de classifica{\c{c}}{\~a}o via an{\'a}lise
explorat{\'o}ria convencional fazendo uso de pertin{\^e}ncia
fuzzy; 2) ado{\c{c}}{\~a}o da t{\'e}cnica de
minera{\c{c}}{\~a}o de dados por {\'a}rvores de decis{\~a}o,
tendo a interven{\c{c}}{\~a}o do int{\'e}rprete na
especifica{\c{c}}{\~a}o do n{\'u}mero m{\'{\i}}nimo de
inst{\^a}ncias por folha; e 3) minera{\c{c}}{\~a}o de dados por
{\'a}rvores de decis{\~a}o sem a interven{\c{c}}{\~a}o direta
do int{\'e}rprete. A avalia{\c{c}}{\~a}o estat{\'{\i}}stica
permitiu selecionar a classifica{\c{c}}{\~a}o de maior
precis{\~a}o para mapear o uso do solo por quadras (IV
Experimento). Os resultados indicaram boa precis{\~a}o nos
mapeamentos realizados, obtendo-se {\'{\i}}ndices Kappa de
0,7618 (Experimento I), 0,7876 (Experimento II), 0,8383
(Experimento III) e 0,7050 (Experimento IV). ABSTRACT: The
products available from the new high spatial resolution satellites
present a high potential for studies and applications in urban
areas. The use of data from these sensor systems opens new
perspectives for the automatic extraction of relevant information
for planning and environmental management. However, for the
efficient information acquisition innovative concepts are
necessary for both, the segmentation and object distinction which
compose the image. In this study satellite images from WorldView-2
were used, which represent on a new Earth Observation concept,
with eight multispectral bands, namely: three bands in the visible
spectrum (Red, Green, Blue), a near-infrared band and the four new
bands Coastal (400-450 nm), Yellow (585-625 nm), Red Edge (705-745
nm) and Near-Infrared-2 (860-1040 nm). The objective of this study
was to evaluate the best attributes and thresholds to map land
use/land cover with WorldView-2 images. The experiment was
performed at a section of western S{\~a}o Paulo Metropolitan
Region. The land cover classification was made in three stages:
1\$^st\$. generation of a classification model through a
conventional exploratory analysis using fuzzy membership;
2\$^nd\$. use of the data mining technique of decision trees
(1\$^th\$ experiment), with the analyst intervention to specify
the minimum amount of instances per node (2\$^th\$ experiment);
3\$^rd\$. data mining by decision trees without the direct
analyst intervention (3\$^th\$ experiment). The statistical
evaluation allowed the selection of the classification with the
highest acuracy to map land use considering urban blocks
(4\$^th\$ experiment). The results indicate a good acuracy of
the maps obtained, with Kappa indices of 0.7618 (1\$^th\$
experiment), 0.7876 (2\$^th\$ experiment), 0.8383 (3\$^th\$
experiment) and 0.7050 (4\$^th\$ experiment).",
committee = "Fonseca, Leila Maria Garcia (presidente) and Kux, Hermann Johann
Heinrich (orientador) and Florenzano, Teresa Gallotti (orientador)
and Almeida, Cl{\'a}udia Maria de and Cruz, Carla Bernadete
Madureira",
copyholder = "SID/SCD",
englishtitle = "Land use and land cover classification with Worldview-2 images
using data mining and object-based analysis",
language = "pt",
pages = "201",
ibi = "8JMKD3MGP7W/3ACT3BL",
url = "http://urlib.net/ibi/8JMKD3MGP7W/3ACT3BL",
targetfile = "publicacao.pdf",
urlaccessdate = "28 abr. 2024"
}