Monday 10 March 2025
A new approach to galaxy classification has been proposed, leveraging the power of machine learning and a unique set of mathematical tools called Zernike moments. These moments are computed for galaxy images, extracting features that can be used to distinguish between different types of galaxies.
The researchers behind this study began by collecting a large dataset of galaxy images from the Galaxy Zoo 2 project. This dataset contains over 11,000 labeled galaxy images, which were then split into two categories: galaxies and non-galaxies. The team computed Zernike moments for each image in the dataset, using these features to train machine learning models.
The first model tested was a support vector machine (SVM), which uses a kernel trick to find the best hyperplane that separates different classes of data. This model achieved an impressive accuracy rate of over 90% when classifying galaxies into spiral, elliptical, and odd objects such as ring galaxies or those with unusual morphology.
The researchers also tested a convolutional neural network (CNN), which uses multiple layers of convolutional and pooling operations to extract features from the galaxy images. This model achieved an accuracy rate of around 85%, slightly lower than the SVM but still impressive considering the complexity of the task.
To further improve the accuracy of the models, the team used transfer learning, which involves pre-training a neural network on a large dataset and then fine-tuning it on the specific problem at hand. This approach significantly improved the performance of both the SVM and CNN, with the CNN achieving an accuracy rate of over 95%.
The use of Zernike moments in this study is particularly noteworthy, as they have been previously used in other fields such as image processing and computer vision. However, their application to galaxy classification has not been explored until now.
The potential applications of this research are vast. By improving the accuracy of galaxy classification, scientists can gain a better understanding of the properties and behavior of different types of galaxies, which could ultimately shed light on the evolution and formation of our universe. Additionally, the use of machine learning and Zernike moments in this study demonstrates the power of interdisciplinary collaboration between astronomy and computer science.
In the future, the researchers plan to apply their methods to even larger datasets, including those containing images from next-generation telescopes such as the Square Kilometre Array (SKA). This will enable them to classify galaxies with unprecedented accuracy and precision, ultimately leading to a deeper understanding of the universe.
Cite this article: “Classifying Galaxies with Machine Learning and Zernike Moments”, The Science Archive, 2025.
Galaxy Classification, Machine Learning, Zernike Moments, Svm, Cnn, Transfer Learning, Astronomy, Computer Science, Image Processing, Galaxy Zoo 2
