Data-parallel methods for fast and deep detection of asteroids on the Umbrella platform: Near-real-time synthetic tracking algorithm for near-Earth objects

Context . Further activities in detecting near-Earth asteroids (NEAs) using the blink method are hampered by the required size of the telescopes. Synthetic tracking (ST) is an effective solution, but the computational demands make operation difficult at survey data rates for fast moving objects. Aims . We aim to show that, through efficient use of hardware, judicious pipeline design and choice of algorithms, ST at survey levels of data rates is possible even for objects with very fast apparent motion, such as NEAs. Methods . We developed a GPU-accelerated ST pipeline, called synthetic tracking on Umbrella (STU), which targets real-time detection of fast NEAs through algorithms designed to make efficient use of data-parallel hardware. STU was developed as part of the Umbrella software suite, which we have expanded to provide an end-to-end data reduction pipeline. Results . We demonstrate the capabilities of the STU pipeline to scan for moving objects faster than the acquisition rate at search radii of 10 arcseconds per minute, with good detection rates on several archival datasets without specific tuning. Conclusions . Our investigation shows that ST is viable for large-scale surveys and that STU may perform such a role.