Researcher · MSc Earth and Space Science · York University
Cleaning up space, one image at a time.
Hello, I'm Connor Humphries, my research focuses on improving the fidelity of synthetic-space-based optical imagery for space situtional awareness (SSA) applications. I work with simulation, image processing, and artificial intelligence to develop synthetic imagery that better represents the physical characteristics and faults of real observations. The goal is to develop imagery that allows algorithms to be trained on synthetic imagery as opposed to needing in-situ imagery which is difficult to obtain.
PCMD
My thesis is focused on the devleopment of the Physics-Constrained Masked Diffusion model for Space Situational Awareness. This project is dual-natured in that it requires development of two avenues and pairing them for the final output. The first is the continued development of the Space-Based Optical Imaging Simulator (SBOIS) and the assumptions it makes within its optical imagery developments in simulating the Fast Auroral Imager (FAI). The second avenue is that of the Diffusion Model, a generative AI model used extensively in image generation. The combination of which shall provide a physics-tied background, with the higher-order noise that is initially difficult to compute.