Automatic meta-data creation for astronomical images

The Virtual Observatory and the field of Semantic Astronomy are attempts to make astronomical meta-data accessible and searchable at a high (semantic) level. However, any structure built on meta-data must have a strategy for dealing with ``meta-crap'' (Doctorow, 2001) or false meta-data. There is overwhelming evidence from the World-Wide Web context that meta-data created independently of its data can never be trusted: humans, whether through malice or mistake, inevitably generate and propagate incorrect meta-data. Only meta-data that can be created or verified by machines, using the data alone, can be trusted.

We have produced a system,, that can generate correct, standards-compliant astrometric meta-data for nearly any astronomical image, using only the image pixels. The input to the system is an image, and the output is the pointing, scale, and rotation of the image on the sky. The system requires no first guess, and works with the information in the image pixels alone. The success rate is above 99.9% for contemporary optical and near-ultraviolet survey data, with no false positives.

Building on this success, we have shown that we can also determine the date an image was taken, to within a few years, by comparing the known proper motions of stars with the positions observed in the image. By using variable stars, we should be able to refine this estimate further. We have also shown that we can determine approximately the bandpass filter through which the image was taken, and calibrate the photometry of the image.

Together, these automated calibration tools obviate the problem of false meta-data (for some kinds of astronomical meta-data), a problem which would otherwise threaten the success of the Virtual Observatory and the Semantic Astronomy effort.

Last modified: Mon Jan 19 17:05:48 GMT 2009