We discuss image segmentation algorithms and additional space considerations for BeaverCube-2, a project under development between the MIT Space Telecommunications, Astronomy, Radiation (STAR) Lab and the Northrop Grumman Corporation that aims to demonstrate the use of an Artificial Intelligence (AI) Computational Accelerator System-on-a-Chip (SoC) on a 3U CubeSat in Low-Earth Orbit (LEO). The processing power afforded by the SoC will allow the use of modern artificial intelligence techniques as part of an Earth observation mission to obtain and process visible and infrared imagery of coastal features.We focus on three algorithms used for cloud segmentation in satellite imagery. These are a luminosity-thresholding method, a random forest method, and an autoencoder-based deep learning method. Our luminosity thresholding method classifies each pixel based on its luminosity and achieved 84% accuracy using 2 MB of memory. Our random forest method contextualizes pixels within a 3 × 3 kernel and classifies them based on the luminosity of each pixel in the kernel -it achieved 90% accuracy, with a memory usage of 700 MB. Finally, our U-Net-based deep learning method achieved 92% accuracy with 1500 MB memory usage, demonstrating modest gains over the two simpler methods, with higher accuracy in snow scenes.

Abstract. Radio occultation (RO) using the global navigation satellite system (GNSS) can be used to infer atmospheric profiles of microwave refractivity in the Earth's atmosphere. GNSS RO data are now assimilated into numerical weather prediction models and used for climate monitoring. New remote sensing applications are being considered that fuse GNSS RO soundings and passive nadir-scanned radiance soundings. Collocating RO soundings and nadir-scanned radiance soundings, however, is computationally expensive, especially as new commercial GNSS RO constellations greatly increase the number of global daily RO soundings. This paper develops a new and efficient technique, called the “rotation–collocation method”, for collocating RO and nadir-scanned radiance soundings in which all soundings are rotated into the time-dependent reference frame in which the nadir sounder's scan pattern is stationary. Collocations with RO soundings are then found when the track of an RO sounding crosses the line corresponding to the nadir sounder's scan pattern. When applied to finding collocations between RO soundings from COSMIC-2, Metop-B-GRAS, and Metop-C-GRAS and the passive microwave (MW) soundings of the Advanced Technology Microwave Sounder (ATMS) on NOAA-20 and Suomi-NPP and the Advanced Microwave Sounding Unit (AMSU-A) on Metop-B and Metop-C for the month of January 2021, the rotation–collocation method proves to be 99.0 % accurate and is hundreds to thousands of times faster than traditional approaches to finding collocations.

No abstract

Abstract. Radio Occultation (RO) using the Global Navigation Satellite Systems (GNSS) can be used to infer atmospheric profiles of microwave refractivity in the Earth’s atmosphere. GNSS RO data are now assimilated into numerical weather prediction models and used for climate monitoring. New remote sensing applications are being considered that fuse GNSS RO soundings and passive nadir-scanned radiance soundings. Collocating RO soundings and nadir-scanned radiance soundings, however, is computationally expensive, especially as new commercial GNSS RO constellations greatly increase the number of global daily RO soundings. This paper develops a new and efficient technique, called the “rotation-collocation-method”, for collocating RO and nadir-scanned radiance soundings in which all soundings are rotated into the time-dependent reference frame in which the nadir sounder’s scan pattern is stationary. Collocations with RO soundings are then found when the track of an RO sounding crosses the line corresponding to the nadir sounder’s scan pattern. When applied to finding collocations between RO soundings from COSMIC-2, Metop-B-GRAS, and Metop-C-GRAS and the passive microwave soundings of ATMS on NOAA-20, Suomi-NPP, and AMSU-A on Metop-B and Metop-C for the month of January, 2021, the rotation- collocation method proves to be 99.0 % accurate and is hundreds to thousands of times faster than traditional approaches to finding collocations.