Available at: https://digitalcommons.calpoly.edu/theses/3288
Date of Award
6-2026
Degree Name
MS in Computer Science
Department/Program
Computer Science
College
College of Engineering
Advisor
John Bellardo
Advisor Department
Computer Science
Advisor College
College of Engineering
Abstract
A star tracker determines spacecraft orientation by photographing the star field, detecting stars in the image, matching them against a catalog, and computing the rotation between observed and cataloged directions. Convolutional neural networks (CNNs) have been proposed as replacements for the detection and centroiding stage, offering improved sub-pixel accuracy and recovering faint stars that classical thresholds lose to stray light and sensor noise. The improvement comes at higher computational cost; the PolySat systemboard targeted in this work lacks the floating-point hardware these networks assume.
This thesis closes the gap between floating-point desktop evaluation and embedded integer deployment. Nine encoder-decoder CNN architectures are quantized to 8-bit integer precision, evaluated end to end against the classical Pyramid identifier and QUEST attitude solver, and the two highest-F1 architectures are deployed on the PolySat systemboard through a C inference engine.
On a 500-image synthetic test set, eight of the nine networks recover detection F1 within 3.4 percentage points of the floating-point baseline; the quantized networks identify stars at 96 to 97 percent against approximately 82 percent for the best-tuned classical detector, with median attitude error between 0.019° and 0.032°. On the systemboard, the deployed model fits within roughly 16 MB of peak resident memory under a tiled-inference strategy.
The results establish a path from floating-point prototype to integer on-target execution for CubeSat star trackers, with on-target inference latency as the principal target for follow-on work.