Morpheus is an artificial intelligence developed by UC Santa Cruz. It is able to automatically analyze astronomical images to identify stars and galaxies…
Astronomical studies generate ever-increasing volumes of data. For example, the soon-to-be-conducted Legacy Survey of Space And Time (LSST) study at the Vera Rubin Observatory in Chile will generate huge amounts of image data.
More than 800 panoramic images will be captured each night by a 3.2 billion pixel camera. The entire visible sky will be recorded twice a week.
This data is very useful for astronomers, and can help them understand how galaxies form and evolve overtime. However, in the face of rapidly increasing the size of astronomical datasets, it has become essential to automate certain tasks of astronomers. For example, it is impossible for human astronomers to classify billions of objects.
This is why researchers at UC Santa Cruz have developed a computer program called Morpheus, capable of analyzing data from astronomical images.
By analyzing pixel-by-pixel data, the program is able to identify and classify all galaxies and stars in data sets generated from astronomical studies.
Morpheus is a framework of Deep Learning. It is based on different artificial intelligence technologies such as image recognition or speech recognition.
The algorithm was driven from data from a 2015 study in which dozens of astronomers classified nearly 10,000 galaxies among Hubble Space Telescope images.
Morpheus: Deep Learning’s first framework specially designed for astronomy
In the past, other astronomers have tried to use Deep Learning to classify galaxies. However, they used existing image recognition algorithms that were fed using images of galaxies to be classified. algorithmes
The difference is that Morpheus was created specifically for the classification of astronomical image data. The model uses the original image data in the standard file format used by astronomers..
Its other advantage is the pixel-level classification. This allows the program to en support very complicated images involving several astronomical forms. .
When Morpheus processes an image of an area of the sky, it generates a new set of images on which all objects are colored according to their morphology. The astronomical objects are separated from the background. Different types of galaxies and stars are identified.
In addition, the algorithm indicates a confidence level for each classification. The program is run on UCSC’s Lux supercomputer and quickly generates pixel-by-pixel analysis of the entire dataset.
This program has been developed over the past two years. However, on May 12, 2020, following the publication of a study in the Astrophysical Journal Supplement Series, the researchers decided to open its code to the public. As a demonstration, you can also view the interactive visualization of Morpheus’ results for the GOODS South study.