When it comes to studying the night sky, astronomers aren’t short of images. There are huge archives of both amateur and professional images taken in the the age before digital imaging. The Harvard College Observatory Astronomical Plate Stacks contain enough images to cover the entire sky 500 times over.
But although the image quality is excellent, the problem is the indexing. When logs go missing or when data has been badly transcribed, it can be almost impossible to work out exactly what appears in an image or when it was taken.
The error rate in many older collections is high enough to make astronomers think twice about using them. And as astronomy moves towards its goal of creating a Virtual Observatory, in which all images are available online in a kind of giant virtual planetarium, a lack of trust in the data is a serious problem.
If, as an astronomer, you’ve been losing sleep over this issue, you can rest easy. David Hogg at New York University and buddies (including the search giant Google), have solved the problem by reducing it to one of image matching.
They take an astrophotograph of dubious provenance and use a computer program called “Blind Date” to look for asterisms (the shapes that constellations make). When they find a match, this immediately locates the image within a part of the sky. But the really cool part of their technique is based on the fact that stars move over time, albeit by tiny amounts. So any small deviation in the location of stars within an image give an immediate time stamp for when the shot was taken.
The team says this technique works for every science-quality image that it has been tested against and 85 per cent of lower quality images. In some cases, it can date images to within a few months.
The plan is to use Blind Date to produce metadata automatically for every image that is entered into the Virtual Observatory, which should reduce errors substantially and also prevent deliberate spoofing of the project.
That should go a long way to restoring trust.
Ref: arxiv.org/abs/0805.0759: Blind Date: Using Proper Motions to Determine the Ages of Historical Images