Research Division Seminar
ParaSOL: data-parallel methods for fast and deep detection of asteroids on the Umbrella platform

Abstract

The conventional (blink) detection techniques require large aperture telescopes for finding the faintest Solar System objects. Nevertheless, unknown objects can be discovered from stacking short exposures across all potential trajectories. This method is called Synthetic Tracking (ST). Due to the large number of potential trajectories, ST is perceived as computationally expensive, thus most surveys still prefer the blink method.

I will present a moving object detection software called Synthetic Tracking on Umbrella (STU), which leverages GPU computation and multiple search-space optimizations for real time synthetic tracking of fast-moving NEOs, even on large, multi-CCD instruments. This software was validated using a large observational archive and multiple mini-surveys, operating in near real-time under realistic conditions. The archive, of more than 100 000 images, was obtained from various telescopes including the 0.25m T025-BD4SB from Bucharest Observatory, the 1.52 m Telescopio Carlos Sánchez, as well as the 2.54m Isaac Newton Telescope. The most active telescopes used for mini-surveys are the 1.6m KASI telescope located in Chile and the aforementioned INT.

The STU detection rate depends on the quality of the input images. For example, on the TCS frames without pre-processing issues no object was missed. Similarly, an analysis on the WFC dataset has shown an 82% detection rate, with most false negatives again coming from pre-processing issues. From the real-time mini-surveys, key examples are 2023 DZ2, a NEO and former Virtual Impactor, discovered by our group in February 2023, and 2024 CW2, a fast-moving NEA (9.5 arcsec/min), reported by us the same day the data was acquired. Even at this apparent motion, the scanning phase of the synthetic tracking operation finished faster (15 min) than the data acquisition (20 min).

Through our work, we have shown that Synthetic Tracking is not as expensive as was once thought and now it can be done at survey speed. For this reason, we expect that Synthetic Tracking will become a foundational method of asteroid discovery and will thus completely displace the blink method and single-image trail detection approaches.