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.

No abstract

No abstract

<p>Small-scale ocean fronts play a significant role in absorbing the excess heat and CO2 generated by climate change, yet their dynamics are not well understood. Existing in-situ and remote sensing measurements of the ocean have inadequate spatial and temporal coverage to map small-scale ocean fronts globally. Additionally, conventional algorithms to generate ocean front maps are computationally intensive and require data with long lead times. We propose machine learning (ML) models to detect temperature and chlorophyll ocean fronts from unprocessed and radiometrically uncorrected satellite im- agery by transfer learning from existing models for edge detection. We use two separate datasets: one based on conventional approaches to ocean front detection, and a second based on human annotated ground truth1. The deep learning front detection approach significantly reduces the resources and overall lead times needed for detecting ocean fronts. The deep learning models are developed with resource-constrained edge compute platforms like CubeSats in mind, as such platforms can address the spatial and temporal coverage challenges. The highest performing models achieve accuracies of 96% and make predictions in milliseconds using unoptimized desktop CPUs and using less than 100 MB of storage; these capabilities are well- suited for CubeSat deployment. </p>