Algorithms for Understanding the Sky


In this talk we will discuss computational issues in making the best use of massive multi-source data from astrophysics. We'll concentrate on two things: First, how to piece together evidence from very many noisy images of the sky as to the locations and orbits of millions of asteroids and minor planet. Second, we will discuss the full pipeline of building Google Sky, and the technologies needed to stitch together and scalably serve deep space surveys to millions of users around the world.

The technical issues for asteroid detection include the use of multiresolution data structures to represent all observed sources on the sky, and dual-tree (and N-tree) algorithms for fast branch-and-bound searches of potential matching asteroid trajectories. The technical issues for creating Google sky center around nonlinearities in color space, and physical pixel space, and stitching together images derived from several different generations of sky imaging technologies from the past forty years.

Work with: Jeremy Kubica, Andy Connolly, Ryan Scranton and many others.

Andrew Moore, Google Pittsburgh

 

Renew SIAM · Contact Us · Site Map · Join SIAM · My Account
Facebook Twitter Youtube