Star Galaxy Image Classification Via Convolutional Neural Networks

This paper gives a review of detailed analysis done on approaches for star galaxy image classification. With the advancements in telescopes, instruments, and computer technology astronomy has become a quite data-rich field of science. From the creation of the first sky atlases to the recent large photometric sky surveys the collected information content is now measured in Terabytes, and Petabytes. Astronomy has morphed to an entirely new stage, the era of huge information, full waveband, and big data, known as Data-driven astronomy. Detecting, classifying, and analyzing sources (star or galaxy) in images in this era requires uniform, efficient, and robust techniques. This research study presents a brief introduction to the issue of star galaxy image classification and then discuss the mechanism, implementation, and performance of various manual and automated methods and technologies used by researchers to address this problem. Furthermore, the performed study analyzes the results of these different approaches and identify the gaps in each to determine how well they adapt to deep space object imagery. The presented review study reveals that deep learning techniques, especially those based on CNN, show remarkable accuracy for these classification problems.