Title of Presentation: Generation of Object-Centric Datasets with Adaptive Sky

Primary (Corresponding) Author: Michael Burl

Organization of Primary Author: Jet Propulsion Lab

Co-Authors: Michael J. Garay, Clare Averill, Benjamin Bornstein, Lukas Mandrake, Justin Ng, Yi Wang

Abstract: Adaptive Sky is an ESTO-funded Advanced Information Systems Technology activity that is developing software to enable multiple sensing assets to be dynamically combined into sensor webs. The ASky feature correspondence toolbox consists of a variety of methods for automatically relating the observations of one instrument at time t to the observations of another instrument at time t'. A key end product from this task will be the ability to generate object-centric datasets in which observations from multiple satellite and in-situ assets are organized not merely by the instrument packaging scheme (e.g., granules, blocks, images, swaths) or by spatial-temporal address (lat, lon, time), but by association with particular physical objects or processes. We have produced an early demonstration of this concept combining selected Adaptive Sky components to identify, track, and reacquire volcanic ash clouds generated by the October 2007 eruption of Bezymianny in Kamchatka. The wide area coverage and high temporal sampling of GOES coupled with ASky tracking capabilities provides a bridge between the less frequent overpasses of NASA's polar orbiting satellites. With this approach, a rich object-centric dataset including measurements from at least five different instruments on four different spacecraft platforms was generated covering a timespan of approximately 30 hours and movement of the ash clouds over 400km from the eruption site.