Image processing and deep learning for space telescope images: an exploratory data analysis approach

Abstract

Exposing patterns in cosmic affairs from space telescope images is performed by a strong methodology configured by Exploratory Data Analysis (EDA) and advanced image processing techniques. This research utilizes machine learning to improve the categorization of stars and galaxies that space telescopes have taken images of, and it uses EDA to uncover hidden phenomena in celestial images. Automated categorization is essential for the purpose of improving relevant research due to the large quantity of astronomical datasets from missions such as the Sloan Digital Sky Survey (SDSS). Having applied contrast enhancement and noise reduction to the images, we use an optimized ResNet50 model for binary classification. In particular, The ResNet50 model which is pre-trained on the ImageNet database was altered to distinguish between stars and galaxies. The results demonstrate how potently image processing and deep learning work together, as seen by the remarkable classification accuracy of around 95%. This method illustrates how machine learning could potentially be used to automate astronomical data analysis, advancing astroinformatics and providing scalable solutions for the space missions to come.