How do you make astrometry fun? Well, you don’t, of course. Astrometry is the precise measurement of the position and motion of astronomical objects. It is not something one takes lightly. It requires skill, exactitude, and lots and lots of math.
Nevertheless, the good folks at Astrometry.net are working on automated methods to do astrometry. They envision sharing their creation with … well, with everyone. They may even succeed in making it fun.
Here’s the game plan. Take any digital image — from a tripod-mounted shot of the sky to a deep view from a major observatory — and automatically determine its precise position and scale, then identify any known objects within it. Such a tool is called a “blind astrometry solver” because, at the start, it knows nothing about the input image. But, just to be greedy, let’s stipulate the solver should still produce correct results even if information provided about the image (so-called metadata) is entirely wrong.
Sam Roweis, a computer scientist at the University of Toronto, and David W Hogg, an astrophysicist at New York University, decided the time was right for such a tool. In 2004, they began receiving developmental support from the U.S. National Science Foundation to apply cutting-edge concepts in machine learning and computer vision to astronomical data sets. This grant kick-started the project.
I stumbled onto Astrometry.net while researching Sloan Digital Sky Survey images for our March issue. So far, the service remains in the alpha-testing phase, available only to selected users; David kindly added me. I then submitted my image of Comet 17P/Holmes from Nov. 11.
A day later, Astrometry.net sent me an e-mail trumpeting its success and giving me a web link to a page with more information. Check it out here.
This is serious stuff, so I asked Astrometry.net’s Dustin Lang, a doctoral student at the University of Toronto, a few questions about the project’s future.
FR: Will Astrometry.net continue as a free service?
DL: Yes. Our goal is to help organize, annotate, and make searchable all the world's astronomical information by providing tools that allow images to be tagged with metadata in a standards-compliant way. We want professional and amateur astronomers to use our software. The more people who use it, the more useful it will become!
And we're not only providing a free web service. We're also releasing our software source code, both so that our users can be confident the software is actually doing what we claim, but also so that it can be adapted for other uses.
FR: I noticed the software reached Version 0.1 in September. What are you currently working on?
DL: Currently we're working on doing more interesting things with the images that are submitted. Right now, the user gets back information about her particular image, but can't really interact with the images others have submitted. We want to provide an interface where amateur and pro astronomers can interact with and explore each other's images in a way that’s fun and encourages collaboration and discovery.
That’s going to require a fair bit of work. We also need to scale up our blind astrometry solver to handle the increased workload we hope this will generate!
FR: What’s the need Astrometry.net is really designed to fill?
DL: There are an enormous number of images of the night sky that are currently difficult to organize because they lack correct metadata. If we can automatically tag these images with a correct astrometry header, this would allow searching and browsing a vast amount of information about the universe. Professional astronomers are very interested in tapping into the images created by amateur astronomers, and our technology allows this.
Other image sources include historical plate archives. For example, the Harvard Observatory has 500,000 astronomical images taken from 1880 to 1990. There are also other images for which incorrect — or no — metadata was recorded. For example, even the Sloan Digital Sky Survey telescope at Apache Point, a sophisticated and highly automated system, had a failure that resulted in perfectly good images with completely incorrect astrometric metadata.
FR: What’s the appeal for amateur astronomers?
DL: Amateur astronomers should find our system interesting in a couple of ways. It can be used to label astronomical objects in images. For example, our alpha-testers have used it to identify nebulae in images. The next phase of the project will allow people to see the images other people have taken of the same part of the sky, which I think will be interesting both in terms of discovering the amazing things the universe has to offer, and also in terms of seeing how other people have made beautiful images. Finally, I imagine some amateur astronomers would be interested in having their images used by researchers, and our system unlocks that potential.
We're particularly interested in facilitating follow-ups on transient events. An amateur astronomer may have taken a very useful photograph of a comet, supernova, or other transient event without even knowing it. Researchers would love to get their hands on these images.