Análise de agrupamentos a partir de parâmetros morfométricos para identificar classes morfológicas de galáxias

• Ano de defesa: 2019
• Autor(a) principal: Gil, Vanessa de Oliveira
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; Brasil; Física; UFSM; Programa de Pós-Graduação em Física; Centro de Ciências Naturais e Exatas
• Palavras-chave em Português: Galáxia; Morfometria; Aprendizado de máquina; Agrupamentos; Galaxy; Morphometry; Machine learning; Clustering analysis; CNPQ::CIENCIAS EXATAS E DA TERRA::FISICA
• Link de acesso: http://repositorio.ufsm.br/handle/1/31395

To understand galaxies processes of formation and evolution, it is essential to identify properties that allow their classification. It is common to perform this type of classification through visual and individual inspection of each image however this method is subjective and limited. It is necessary to seek for an automatic method based on the morphology quantification to find classes of galaxies. It is proposed an automated separation through clustering analysis for objects in the space of morphometric parameters as measured by Morfometryka. This method aims to find groups of strongly correlated observations that indicates the same physical processes of formation and evolution. The study object corresponds to a sample of 1808 images of EFIGI survey with galaxies of all morphological types. To achieve this goal, a set of parameters based on their morphometry were selected to input in the clustering algorithm: concentration C1, asymmetry A1, smoothness S3, Gini Coefficient G, Entropy H, Spirality , and in Sérsic Law In, Rn and n; axis ratio q and integrated brightness LT . The Ward clustering algorithm is based on hierarchy. With this method three large groups of galaxies were found in EFIGI and 12 subgroups with peculiar characteristics that help in the formation of three main classes: spheroids, spirals and irregulars. To identify as much information as possible from these groups, a principal component analysis was performed to determine which variables have the greatest influence on the construction of each class of galaxies found. The inspection of these results through the discriminant analysis provided models that statistically describe the objects within the obtained groups.