AAS 207th Meeting, 8-12 January 2006
Session 26 LSST
Poster, Monday, 9:20am-7:00pm, January 9, 2006, Exhibit Hall

Previous   |   Session 26   |   Next  |   Author Index   |   Block Schedule

[26.32] Efficiently Tracking Moving Sources in the LSST

J. Kubica (CMU), T. Axelrod, K. Barnard (U. Arizona), A. Connolly (U. Pittsburgh), L. Denneau (UH), A. Efrat (U. Arizona), J. Heasley, R. Jedicke (UH), B. Moon (U. Arizona), A. Moore (CMU), S. Morris, P. Rao (U. Arizona)

The LSST will survey the sky with a cadence of several visits per month spaced by approximately a week. This time sampling enables a detailed census of Solar System objects ranging from over a million Main Belt Asteroids, to 20,000 Trans Neptunian Objects and even potentially hazardous asteroids. The challenge in identifying all potential asteroid associations is that it can be very computationally expensive (due to the many potential tracks that must be tested in order to isolate true orbits).

To address these issue, we have developed new approaches for tracking asteroids and for recovering orbits from isolated observations that use spatial structures in order to ignore ``obviously'' bad candidate sets.

The goal here is to reduce the number of PROPOSED associations that must be tested, not just to reduce the cost of a test. To this end, we have developed a new class of search algorithms based on traditional spatial data structures (e.g. KD-trees) that use a variable number of tree nodes to capture information about the current state of a search. These algorithms are able to exploit simultaneously all available observations in order to limit the number of candidate asteroid tracks. Initial applications show that these approaches scale to the size of the LSST domain, can recover 95% of asteroids with a small false positive rate and have the potential to provide almost real time identification of moving sources.

We validate and improve orbits computed from the linkages by matching future and prior observations to proposed orbits. Here we consider the probability that eachobservation should be paired with each hypothesized orbit in its vicinary, and maximize the overall association likelihood using bipartite graph matching. This gives a global solution that enforces the one-to-one constraint. This approach also provides for control of outliers, and we propose some strategies to allow mismatches consistent with estimated probabilities that the items do not have a match.

Previous   |   Session 26   |   Next

Bulletin of the American Astronomical Society, 37 #4
© 2005. The American Astronomical Soceity.