@MastersThesis{Rocha:2023:PrImSa,
author = "Rocha, Brenda Oliveira",
title = "Processamento de imagens dos sat{\'e}lites brasileiros CBERS-4,
CBERS-4a e Amazonia-1 para respostas r{\'a}pidas a desastres",
school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
year = "2023",
address = "S{\~a}o Jos{\'e} dos Campos",
month = "2023-03-21",
keywords = "desastres, deslizamentos, inunda{\c{c}}{\~o}es, sat{\'e}lites
brasileiros, processamento de imagens, disasters, landslides,
floods, brazilian satellites, image processing.",
abstract = "Dentre a pluralidade de desafios que surgem para a gest{\~a}o de
desastres naturais, a fase de resposta p{\'o}s-desastre pode ser
considerada a mais desafiadora, tendo em vista a necessidade do
fornecimento r{\'a}pido de informa{\c{c}}{\~o}es que auxiliem
nesse processo. Considerando as muitas vantagens do Sensoriamento
Remoto (SR), as imagens de sat{\'e}lite podem contribuir para a
an{\'a}lise da extens{\~a}o das ocorr{\^e}ncias e a
identifica{\c{c}}{\~a}o das {\'a}reas mais afetadas,
atrav{\'e}s da utiliza{\c{c}}{\~a}o de t{\'e}cnicas de
Processamento Digital de Imagens (PDI) que revelam {\'a}reas de
interesse. O International Charter Space and Major Disasters
(Carta) {\'e} a principal coopera{\c{c}}{\~a}o mundial entre
ag{\^e}ncias espaciais para o fornecimento gratuito de dados de
emerg{\^e}ncia. A coopera{\c{c}}{\~a}o conta com a
contribui{\c{c}}{\~a}o do Brasil no processo de resposta aos
chamados, quando a ocorr{\^e}ncia se encontra dentro da {\'a}rea
de cobertura dos sat{\'e}lites brasileiros. Dando
import{\^a}ncia tanto para os chamados da Carta quanto para
outras eventuais solicita{\c{c}}{\~o}es de emerg{\^e}ncia, a
presente pesquisa teve por objetivo utilizar dados oriundos dos
sat{\'e}lites brasileiros e sistematizar t{\'e}cnicas de PDI
para o apoio {\`a} gest{\~a}o de desastres do tipo deslizamentos
de terra e inunda{\c{c}}{\~o}es regionais. A
minera{\c{c}}{\~a}o de dados foi adotada para extrair os
principais atributos obtidos a partir das t{\'e}cnicas de PDI
aplicadas aos produtos dos sensores nacionais (WFI/CBERS-4,
WFI/AMAZONIA-1, MUX/CBERS-4A e WPM/CBERS-4A). Com o apoio do
algoritmo Random Forest (RF), foi realizada uma
classifica{\c{c}}{\~a}o supervisionada onde os tr{\^e}s
atributos de maior relev{\^a}ncia para a
classifica{\c{c}}{\~a}o foram combinados no espa{\c{c}}o de
cores RGB para a identifica{\c{c}}{\~a}o r{\'a}pida das
{\'a}reas atingidas. A metodologia foi testada em quatro casos de
estudo, os quais s{\~a}o: 1) deslizamentos ocorridos no
in{\'{\i}}cio de 2022 em Petr{\'o}polis {{(RJ);}} 2)
deslizamentos ocorridos em maio de 2022 em Recife {{(PE);}} e as
inunda{\c{c}}{\~o}es regionais ocorridas em: 3) tr{\^e}s
grandes prov{\'{\i}}ncias do Paquist{\~a}o em agosto de
{{2022;}} e 4) inunda{\c{c}}{\~a}o nos munic{\'{\i}}pios de
Itamaraju e Prado (BA) em 2021. O mapeamento realizado pela Carta
em cada caso de estudo foram utilizados como refer{\^e}ncia
{\`a} avalia{\c{c}}{\~a}o quantitativa e qualitativa das
composi{\c{c}}{\~o}es finais sugeridas. Para os casos de
deslizamentos, a composi{\c{c}}{\~a}o proposta para
Petr{\'o}polis foi (R=CP3, G=SAVI, B=HUE), com uma Acur{\'a}cia
Global (AG) de 81,82%, obtida a partir dos dados do MUX/CBERS-4A.
Para o caso de Recife, a composi{\c{c}}{\~a}o proposta foi
(R=CP3, G=NDWI, B=CP4), com uma AG de 76,70%, com base nos dados
do WPM/CBERS/4A. Sobre as inunda{\c{c}}{\~o}es regionais, a
composi{\c{c}}{\~a}o sugerida para o Paquist{\~a}o foi (R=NIR,
G=CP1, B=CP2), com uma AG de 88,33%, utilizando imagens do
WFI/CBERS-4. Para o caso das cidades da Bahia, a
composi{\c{c}}{\~a}o foi (R=NIR, G=CP1, B=EVI), com uma AG de
98,08%, alcan{\c{c}}ada a partir dos dados do WFI/AMAZONIA-1. A
terceira componente principal (CP3) apresentou resultados
relevantes no caso dos deslizamentos, assim como a banda do NIR
para o caso das inunda{\c{c}}{\~o}es regionais. Todas as
{\'a}reas de interesse puderam ser evidenciadas nas
composi{\c{c}}{\~o}es sugeridas, com a observa{\c{c}}{\~a}o de
um melhor contraste entre os alvos sem a necessidade de
aplica{\c{c}}{\~a}o de limiares. ABSTRACT: Among the plurality
of challenges that arise for the management of natural disasters,
the post-disaster response phase can be considered the most
challenging, in view of the need to quickly provide information
that helps in this process. Considering the many advantages of
Remote Sensing (RS), satellite images can contribute to the
analysis of the extent of occurrences and the identification of
the most affected areas, through the use of Digital Image
Processing (DIP) techniques that reveal areas of interest. The
International Charter Space and Major Disasters (Charter) is the
world's leading cooperation between space agencies for the free
provision of emergency data. The cooperation relies on Brazil's
contribution in the process of responding to calls, when the
occurrence is within the coverage area of Brazilian satellites.
Giving importance to both the Charter calls and other possible
emergency requests, the present research aimed to use data from
Brazilian satellites and systematize DIP techniques to support the
management of disasters such as landslides and regional floods.
The data mining technique was adopted to extract the main
attributes obtained from the PDI techniques applied to national
sensor products (WFI/CBERS-4, WFI/AMAZONIA-1, MUX/CBERS-4A and
WPM/CBERS-4A). With the support of the Random Forest (RF)
algorithm, a supervised classification was performed where the
three most relevant attributes for the classification were
combined in the RGB color space for the quick identification of
the affected areas. The methodology was tested in four case
studies, which are: 1) landslides that occurred in early 2022 in
Petr{\'o}polis {{(RJ);}} 2) landslides that occurred in May 2022
in Recife {{(PE);}} and the regional floods that occurred in: 3)
three major provinces of Pakistan in August {{2022;}} and 4)
flooding in the municipalities of Itamaraju and Prado (BA) in
2021. The mapping carried out by the Charter in each case study
was used as a reference for the quantitative and qualitative
evaluation of the suggested final compositions. For cases of
landslides, the composition proposed for Petr{\'o}polis was
(R=CP3, G=SAVI, B=HUE), with a Global Accuracy (GA) of 81.82%,
obtained from MUX/CBERS-4A data. For the case of Recife, the
proposed composition was (R=CP3, G=NDWI, B=CP4), with an GA of
76.70%, based on WPM/CBERS/4A data. On regional flooding, the
suggested composition for Pakistan was (R=NIR, G=CP1, B=CP2), with
an GA of 88.33%, using images from WFI/CBERS-4. For the case of
the cities in Bahia, the composition was (R=NIR, G=CP1, B=EVI),
with an GA of 98.08%, obtained from WFI/AMAZONIA-1 data. The third
principal component (CP3) presented relevant results in the case
of landslides, as well as the NIR band for the case of regional
floods. All areas of interest could be evidenced in the suggested
compositions, with the observation of a better contrast between
the targets without the need to apply thresholds.",
committee = "Renn{\'o}, Camilo Daleles (presidente) and K{\"o}rting, Thales
Sehn (orientador) and Namikawa, Laercio Massaru and Ferreira,
Antonio Geraldo",
englishtitle = "Image processing from the brazilian satellites CBERS-4, CBERS-4a
and Amazonia-1 for quick responses to",
language = "pt",
pages = "103",
ibi = "8JMKD3MGP3W34T/48Q9KJE",
url = "http://urlib.net/ibi/8JMKD3MGP3W34T/48Q9KJE",
targetfile = "publicacao.pdf",
urlaccessdate = "24 abr. 2024"
}