A Novel Deep Learning/Machine Learning Hybrid Technique for Automatic Classification of Nebulae

Overview

There are believed to be ~20,000 nebulae in the Milky Way Galaxy; however, humans have only cataloged ~1,800 of them even though we have gathered 1.3 million observations. Resources like the Hubble telescope can automatically explore space and discover new artifacts. Still, their classification is a human skill, which ultimately is interminable and subject to human error. The importance of classifying nebulas cannot be stressed enough. Studying the chemical composition of a nebula can help us understand the material of the original star. My research of nebulae classification aims to make the process of discovering and classifying new nebulae faster and more accurate using a hybrid of Deep Learning and Machine Learning techniques. Nebulae can be classified into five different categories: planetary, supernova remnants, emission, reflection, and dark. Using a dataset primarily of images from the European Space Agency, I experimented with a range of artificial intelligence techniques and determined the deep learning network/machine learning algorithms that produced the best results. Some experiments included converting the images to black and white, featurization, and dropping specific categories. The main conclusions reached from my research were that the AI was not dependent on color to classify nebulae, dropping specific categories increased accuracy and that featurization is the most effective technique to classify nebulae accurately. The discovery of new astronomical bodies is already mechanized. Making the classification of nebulae automated will help us discover, identify, and classify these marvels much faster and accurately.

Competition history

• JSHS 2022 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.