Computer vision and machine-learning method for the detection of low-surface brightness galaxies in the Fornax Cluster
Loading...
Date
2024
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Detecting the faint luminosity in Low Surface Brightness Galaxies (LSBGs) poses significant challenges, primarily due to sky brightness and contamination from brighter sources while separating LSBGs from the background. Despite these challenges, the study of LSBGs holds great potential to advance our understanding of various fields, including cosmology, galaxy formation, evolution, and the characteristics of galaxy clusters. The primary goal of this study is to develop an automated code capable of effectively detecting LSBGs, including the more diffuse LSBGs that are only detectable through visual search. The initial focus is on the Fornax cluster of galaxies, with the possibility of extension to other galaxy clusters. The purpose is to significantly contribute to advancing research in LSBGs and its implications for broader astronomical studies. We have created an automated code that successfully detects LSBGs in digital images at a reasonable processing speed. We have incorporated an innovative algorithm to separate LSBGs from the background using a dynamic background kernel and threshold applied to image segments to achieve this. We have also implemented a bilateral filter that identifies the most diffuse LSBGs and preserves morphology, ensuring precise identification and classification. Additionally, we have developed and trained a One-Class Support Vector Machine (SVM) classifier using a gold sample of 143 LSBGs, resulting in a classifier with a low rate of false positives. The implemented code has successfully detected LSBGs, showcasing its ability to address the challenges associated with identifying the faint luminosity in these galaxies, even in the presence of brighter sources. The integrated algorithm has significantly improved the accuracy and efficiency of the detection process, allowing for the identification of a substantial number of LSBG candidates. Specifically, in the Fornax Cluster, our algorithm successfully identified 31,295 LSBG candidates, as documented in the comprehensive catalog available at GitHub Repository: \url{https://github.com/Alevhf/LSB_candidates/blob/main/Catalog_result.csv#L19304}.
Description
Tesis (Magíster en Astrofísica)--Pontificia Universidad Católica de Chile, 2024.