@InProceedings{GenovezPoMaToBeSiMi:2023:TrApO,
               author = "Genovez, Patr{\'{\i}}cia Carneiro and Ponte, Francisco 
                         F{\'a}bio de Ara{\'u}jo and Matias, {\'{\I}}talo de Oliveira 
                         and Torres, Sarah Barr{\'o}n and Beisl, Carlos Henrique and 
                         Silva, Gil M{\'a}rco Avellino and Miranda, Fernando Pellon de",
          affiliation = "{Pontif{\'{\i}}cia Universidade Cat{\'o}lica do Rio de Janeiro 
                         (PUC-Rio)} and {Pontif{\'{\i}}cia Universidade Cat{\'o}lica do 
                         Rio de Janeiro (PUC-Rio)} and {Pontif{\'{\i}}cia Universidade 
                         Cat{\'o}lica do Rio de Janeiro (PUC-Rio)} and 
                         {Pontif{\'{\i}}cia Universidade Cat{\'o}lica do Rio de Janeiro 
                         (PUC-Rio)} and {Geoespa{c{c}}o} and Petrobras and Petrobras",
                title = "Desenvolvimento e aplica{\c{c}}{\~a}o de modelos preditivos para 
                         distinguir seepage slicks oil spills em imagens SAR da 
                         superf{\'{\i}}cie do mar: transfer{\^e}ncia de aprendizagem 
                         entre o Golfo do M{\'e}xico e a margem continental brasileira",
            booktitle = "Anais...",
                 year = "2023",
               editor = "Gherardi, Douglas Francisco Marcolino and Arag{\~a}o, Luiz 
                         Eduardo Oliveira e Cruz de and Sanches, Ieda DelArco",
                pages = "e155920",
         organization = "Simp{\'o}sio Brasileiro de Sensoriamento Remoto, 20. (SBSR)",
            publisher = "Instituto Nacional de Pesquisas Espaciais (INPE)",
              address = "S{\~a}o Jos{\'e} dos Campos",
             keywords = "Radar de Abertura Sint{\'e}tica, Machine Learning, Transfer 
                         Learning, Exsuda{\c{c}}{\~a}o de {\'O}leo, Derrame de 
                         {\'O}leo, Synthetic Aperture Radar, Machine Learning, Transfer 
                         Learning, Seepage Slick, Oil Spill.",
             abstract = "Manchas de {\'o}leo naturais ou antr{\'o}picas induzem a 
                         atenua{\c{c}}{\~a}o da rugosidade da superf{\'{\i}}cie do mar, 
                         sendo igualmente detectadas como alvos escuros por Radares de 
                         Abertura Sint{\'e}tica (SAR). No Golfo do M{\'e}xico (GoM), onde 
                         seepage slicks e oil spills podem ocorrer simultaneamente, 
                         distinguir a origem das manchas de {\'o}leo (OMO) usando SAR 
                         {\'e} desafiador. Modelos preditivos para 
                         identifica{\c{c}}{\~a}o da OMO no GoM foram desenvolvidos 
                         utilizando 26 atributos geom{\'e}tricos, extra{\'{\i}}dos de 
                         6.279 manchas de {\'o}leo validadas. Os modelos GoM treinados e 
                         testados com algoritmos de Machine Learning alcan{\c{c}}aram 
                         precis{\~a}o m{\'a}xima de 75%. De forma in{\'e}dita, estes 
                         modelos foram aplicados para prever amostras desconhecidas na 
                         Margem Continental Brasileira utilizando Transfer Learning. Os 
                         resultados demonstraram a capacidade de generaliza{\c{c}}{\~a}o 
                         dos modelos GoM atingindo 87% de precis{\~a}o empregando 
                         sat{\'e}lites semelhantes. Predi{\c{c}}{\~o}es autom{\'a}ticas 
                         agregam confian{\c{c}}a {\`a} an{\'a}lise dos int{\'e}rpretes, 
                         minimizando riscos geol{\'o}gicos para gera{\c{c}}{\~a}o e 
                         migra{\c{c}}{\~a}o de {\'o}leo em novas fronteiras 
                         explorat{\'o}rias offshore. ABSTRACT: Natural or anthropic oil 
                         slicks induce the sea surface roughness attenuation, being 
                         similarly detected as dark spots by Synthetic Aperture Radars 
                         (SAR). Thereby, in the Gulf of Mexico (GoM), where seepage slicks 
                         and oil spills can occur simultaneously, distinguishing the oil 
                         slick source (OSS) using SAR is challenging. A database with 26 
                         geometric features, extracted for 6,279 validated oil slicks, was 
                         used to develop predictive models for OSS identification in the 
                         GoM. A Machine Learning processing chain was implemented to train 
                         and test the GoM models achieving maximum accuracy around 75%. 
                         These models were first-ever applied to predict unknown samples in 
                         the Brazilian Continental Margin employing Transfer Learning. 
                         Results demonstrated the generalization capacity of the GoM 
                         models, achieving 87 % of accuracy when using similar satellites. 
                         Automatic predictions add confidence to the interpreters analysis, 
                         minimizing inherent risks regarding oil generation and migration 
                         in new offshore exploratory frontiers.",
  conference-location = "Florian{\'o}polis",
      conference-year = "02-05 abril 2023",
                 isbn = "978-65-89159-04-9",
             language = "pt",
         organisation = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                  ibi = "8JMKD3MGP6W34M/48UQ695",
                  url = "http://urlib.net/ibi/8JMKD3MGP6W34M/48UQ695",
           targetfile = "155920.pdf",
                 type = "Intelig{\^e}ncia Artificial para Observa{\c{c}}{\~a}o da 
                         Terra",
        urlaccessdate = "06 maio 2024"
}