Dino Bektešević (U. Washington): "Kernel Based Moving Object Detection"

Kernel Based Moving Object (KBMOD) detection is an algorithm that searches overlapping images for moving objects below the single image detection limit. By guessing a trajectory and integrating the signal along it, the limiting detection magnitude can be lowered at a cost of computational resources. Key to the approach is the use of GPU acceleration which enables testing of 10^12 - 10^13 hypothetical trajectories per minute. With a decade long history at University of Washington, KBMOD has been used to find trans-Neptunian Objects (TNOs) in High Cadence Transient Survey (HiTS), DECam Ecliptic Exploration Project (DEEP) data and has, in the last year, been preparing to run searches on the scale of the upcoming Legacy Survey of Space and Time (LSST). Observations of TNOs provide key insights for both the composition and the dynamic evolution of the early Solar System. LSST expects to find 40,000 new TNOs, an order of magnitude more than currently known. Similarly, our estimates show that a decrease of the r band detection limiting magnitude from the expected 24.7 to 26.1, equivalent to searching 3 months of LSST data with KBMOD, could yield approximately 8 times more TNOs.