Every night, thousands of amateur astronomers in their backyards point digital cameras and telescopes at the same bits of starry sky that professional scientists scan from mountaintop domes. Although both groups collect thousands of images, they rarely use one another’s results. While amateurs are more interested in aesthetics, professionals need hard numbers.
In a first step toward bridging this divide, a team of astronomers and computer scientists has created pattern-recognition software that may provide an easy way for the two groups to collaborate by making their astronomical images equally searchable. The Web-based application, scheduled for a beta release in early 2008 at Astrometry.net, can analyze nearly any field of stars and, based on the particular geometric relationships of the stars, determine exactly which part of the sky the photo captures. The terrestrial equivalent would be a program that could pinpoint the latitude and longitude of your house from an aerial photograph of your street.
”The vast majority of astronomical data is in disarray,” explains David Hogg, a New York University astronomer in New York City who three years ago conceived of the project with his high school classmate Sam Roweis, now a computer science professor at the University of Toronto. Hogg points to boxes of magnetic tape on his office shelf containing digitized images and explains that it would be easier to apply for new telescope time and re-collect the data than to get what he needs from the tapes. As another example, he notes that Harvard University has one of the world’s largest archives of astronomical images—nearly half a million plates dating from the era before digital imaging—but the handwritten logs make them hard to search, so many just gather dust.
Automatically determining an image’s location in the sky provides the first step toward making both forgotten professional data and images from amateurs searchable and standardized. ”Professional astronomers are great with taking pictures of the sky,” says Roweis, but comprehensive surveys happen only once or twice a decade. ”Amateur astronomers, on the other hand, take pictures every day,” which can be valuable for studying fast-changing astronomical events, he says.
Astrometry.net’s search software begins its analysis by looking for the brightest stars in the image and then uses sets of four such stars to draw four-sided shapes that Hogg and his colleagues call quads. Each quad is like a fingerprint for a particular part of the sky. But because there are so many stars in the sky—a few thousand visible to the naked eye alone, and billions visible to telescopes—many of these fingerprints look similar.
Rather than trying to find a perfect match, the program looks at many possible matches in a database of more than a billion stars, according to Dustin Lang, a University of Toronto Ph.D. student who, together with fellow student Keir Mierle, wrote most of the code. For each matching quad, the computer compares surrounding stars in the image to those predicted by information in the database and reports a successful match only when the stars’ positions correspond with little discrepancy.
To attract more amateur interest, Hogg and his team hope to combine their program with online virtual planetariums such as Google Earth’s Sky feature, Microsoft’s upcoming World-Wide Telescope, and an open-source project called Stellarium.
Once launched, Astrometry.net will allow amateurs to superimpose matching images from the Hubble Space Telescope and other professional sources on top of their own photos, says Hogg, or to identify all the stars and constellations in a backyard snapshot. A small group of testers has already found new ways to use the open-source software, and Hogg and Roweis plan to make it available to the public as soon as they secure funding to support more users.