StereoBit: Onboard Intelligence for a Stereo Winds Constellation

Abstract:

Stereo imaging is a powerful technique to track atmospheric features in motion and derive winds with geometric height assignments. We have demonstrated the advantages of multi-platform stereo methods using NASA’s MISR and MODIS instruments observing in tandem with Geostationary (GEO) satellites such as GOES-16/17, and Himawari-8. Such methods are well suited for a Low Earth Orbital (LEO) constellation of small satellites carrying compact cameras that observe cooperatively with each other or with GEO satellites. Onboard science processing is required to enable the downlink of science data on a 24x7 basis and to achieve low latency global coverage. StereoBit, supported by NASA/ESTO’s Advanced Information Systems Technology (AIST) program, is an application for the SpaceCube-Mini family of hybrid onboard processors with size, weight, and power suitable for CubeSats. StereoBit calibrates, navigates, and applies Structure from Motion (SfM) algorithms to measure disparities between images of clouds being tracked by a camera such as the multi-spectral, multi-angle Compact Midwave Imaging System (CMIS). StereoBit extracts the science content onboard from the raw data and thereby significantly reduces downlink data volume. We exploit SpaceCube’s Field Programmable Gate Arrays (FPGAs) to parallelize and accelerate StereoBit. Our paper discusses the advantages of stereo winds, presents results from our StereoBit prototype using flight-derived proxy data, and lays out a vision for future constellation missions enabled by embedded onboard intelligence and cooperative observing.