RING-Id: uma ferramenta robusta para a identificação automática de galáxias peculiares do tipo aneladas

Abstract

In this Dissertation, is presented a robust tool to identify peculiar ring galaxies automatically, called (RING-Id) that uses Neural Artificial Net and the SURF Algorithm by means the features extraction and the LIRe. As a data set, was used the catalogues of Faúndez-Abans & Oliveira-Abans (1998) and Moiseev et al. (2011) and DSS images. Due to the quality of those images and to have a good statistical sample of galaxies, at the moment the tool identify two categories of ring galaxies, the elliptical and the polar ring. The training and classification samples have 124 and 53 images, respectively. In the training phase, the accuracy and the F-Score were 98,0% and 98,6%, respectively. On the other hand, for the classification phase, those values were 81,0% and 73,0%, respectively. The tool contains two modules: i-) training, ii-) classification and visualisation. It can also be adapted to be used in any Large Astronomical Survey image to identify any other peculiar galaxy category.