Session
Advanced Technologies Research & Academia 1
Location
Salt Palace Convention Center, Salt Lake City, UT
Abstract
The inaugural 1U CubeSat mission of Universidad Panamericana represents a shift in academic remote sensing, prioritizing ground-segment intelligence to overcome the physical and computational constraints of a 10-centimeter cubic form factor. Given strict power limitations, the satellite utilizes a low-cost, space-qualified XCAM C3D imaging system providing moderate-quality 1.3MP imagery with a ground sample distance (GSD) of approximately 360 meters at an altitude of 550 kilometers. To transform these raw frames into high-value Earth Observation products, this paper details an automated ground station framework hosting a modular image processing pipeline. The pipeline combines traditional image processing—including histogram equalization and edge enhancement—with automated decision logic used to select specific deep learning restoration models. To optimize ground computing resources, these heavy restoration models are applied conditionally. A frequency-domain analysis activates a DeblurGAN module for motion blur recovery when the average Fast Fourier Transform (FFT) magnitude of the image falls below 137. Simultaneously, a lightweight semantic segmentation layer triggers a CloudGAN module for contextual cloud removal only when regional cloud cover exceeds 40%. To ensure that all stored mission images achieve a standardized, high-resolution format, an Enhanced Deep Super-Resolution (EDSR) network is applied to every processed frame. Comparative analysis against alternative models demonstrates that while transformer options provide higher structural similarity, EDSR achieves the highest numerical fidelity with a Peak Signal-to-Noise Ratio (PSNR) of 24.86 dB within a 10-minute processing window. This operational framework provides a scalable blueprint to expand the data value of low-cost satellite missions without requiring on-orbit hardware changes.
Document Type
Event
An AI-Enabled Ground Processing Pipeline for Image Restoration and Super-Resolution in 1U CubeSat Missions
Salt Palace Convention Center, Salt Lake City, UT
The inaugural 1U CubeSat mission of Universidad Panamericana represents a shift in academic remote sensing, prioritizing ground-segment intelligence to overcome the physical and computational constraints of a 10-centimeter cubic form factor. Given strict power limitations, the satellite utilizes a low-cost, space-qualified XCAM C3D imaging system providing moderate-quality 1.3MP imagery with a ground sample distance (GSD) of approximately 360 meters at an altitude of 550 kilometers. To transform these raw frames into high-value Earth Observation products, this paper details an automated ground station framework hosting a modular image processing pipeline. The pipeline combines traditional image processing—including histogram equalization and edge enhancement—with automated decision logic used to select specific deep learning restoration models. To optimize ground computing resources, these heavy restoration models are applied conditionally. A frequency-domain analysis activates a DeblurGAN module for motion blur recovery when the average Fast Fourier Transform (FFT) magnitude of the image falls below 137. Simultaneously, a lightweight semantic segmentation layer triggers a CloudGAN module for contextual cloud removal only when regional cloud cover exceeds 40%. To ensure that all stored mission images achieve a standardized, high-resolution format, an Enhanced Deep Super-Resolution (EDSR) network is applied to every processed frame. Comparative analysis against alternative models demonstrates that while transformer options provide higher structural similarity, EDSR achieves the highest numerical fidelity with a Peak Signal-to-Noise Ratio (PSNR) of 24.86 dB within a 10-minute processing window. This operational framework provides a scalable blueprint to expand the data value of low-cost satellite missions without requiring on-orbit hardware changes.
