AI and Space: A Match Made in Heaven
Category: AI insights
Published date: 25.06.2020
Read time: 7 min
Over the course of human history, there has always been a fascination with stars, planets, comets, asteroids, and many other space objects. Thanks to recent discoveries powered by modern technology, we know a lot more about what goes on in our solar system and beyond. AI is assisting researchers to make more progress in identifying space bodies and learning about their movements and inner-workings. Today we would like to tell you about some of the interesting new AI-powered technologies astronomers, astrophysicists, and other researchers are using to make additional progress into space exploration.
Detecting and Classifying Galaxies
Beyond our own galaxy, there are countless other ones each holding interesting new secrets about the workings of planets and other space objects contained inside. Since there are so many galaxies out there, it is not possible for researchers to detect and classify each one. A new project out of the University of California, Santa Cruz, called Morpheus is using AI technology to detect and classify galaxies using pixel-level morphological classification. This allows the technology to avoid false positives properly to identify each object in an image as a galaxy.
In order to train the machine learning algorithm, researchers took data from a 2015 study where astronomers identified about 10,000 galaxies in the images detected by the Hubble Space Telescope. While Morpheus is currently only able to detect galaxies, if the project were to be expanded to identify stars located within those galaxies or anything else for that matter, similar data annotation by humans would also be necessary. Even if they would not be able to find already labeled data sets, they could hire astronomers or other competent individuals to annotate the data.
On that note, this project requires semantic segmentation, which is the most precise form of data annotation out there. It requires linking each pixel in an image to a class label and that is exactly what we are seeing here. Such a level of precision is very important because the astronomical object needs to be separated from the background to identify the point source. This is necessary because not all galaxies are the same and all of the galaxy types can be further subdivided into categories for further study.
Looking for Extraterrestrial Life Forms
While many researchers are suspecting that foreign life exists in various ends of the solar system, no definitive proof has ever been found. Due to recent sci-fi movies, we have our own imagination and beliefs in terms of the way such life forms communicate. Thanks to research from the Search for Extraterrestrial Intelligence Institute (SETI), we are finding out more information about possible life on other planets. They are using AI technology to scan anomalies that may be by-products of a technologically advanced civilization going about its business. Their goal is to create a power anomaly engine that can detect data values that correlate to unusual patterns when compared to the ground truth data. For example, in 2001, there was a powerful burst of radio energy that lasted no more than 5 milliseconds. This was detected only six years later but the new engine that is being worked on by SETI can detect this a lot sooner.
Using AI to Observe Weather and Environmental Patterns
Satellites are producing thousands and thousands of Earth images every day which amounts to 150 terabytes of data every day. Some of these images capture severe weather patterns and allow meteorologists and emergency services to send out severe weather advisories a lot more quickly. Researchers from Penn State University and Accuweather Inc. are working together to track the shapes and movements of clouds that could be a sign of severe storms. Such technology is also used to identify hurricanes earlier and even atmospheric conditions that are suitable for the potential formation of a hurricane. For example, in the Atlantic Ocean, hurricanes can form from sub-Saharan thunderstorms if the atmospheric conditions allow for it.
Thanks to the satellite images, researchers can learn about the formation of hurricanes and their potential travel paths. All of the images are fed into supercomputers that analyze all of the data and identify dangerous weather patterns. There are many supercomputer centers across the United States, EU, and other parts of the globe where all of the data can be analyzed. Even though these supercomputers are all cutting-edge technologies, there is still a human element in play since the machine learning algorithms that power them need to be trained. Just like in the example earlier where a system needed to be trained to identify galaxies, these supercomputers, and subsequent technologies will need to be trained to identify storms, hurricanes, and other severe weather patterns.
Mindy Support is Here to Assist Aerospace Researchers
Some of the projects that we looked at require a lot of data annotation to train machines to recognize objects and patterns. This is a lot of time-consuming yet necessary work that needs to be done in order for the system to function properly. Mindy Support can take this part of the project off the shoulder of your internal team and annotate all of the data you need with unmatched quality. We have a rigorous QA process in place to make sure everything is done correctly and we have extensive experience actualizing large scale projects. Regardless of the volume of data you need to be annotated, we can assemble even the most sizeable team for you quickly without compromising the quality of the annotations.
Posted by Il’ya Dudkin