Robust onboard space object detection using tri-part cnn transformer model
TriCCOT: Tri-part Convolutional Conformal Transformer for Onboard Space Object Detection
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Detecting objects in images taken from space is hard because computers on satellites have limited power and the images can be blurry or noisy. The authors created TriCCOT, which uses a combination of different techniques: a fast method to find likely object areas, a way to confidently guess object boundaries, and a new kind of attention mechanism that works well on special hardware. This approach keeps detection accurate even when images are less clear, and it runs efficiently on space-grade chips without needing hardware changes.
What this means in practice
- •For satellite system engineers: Implement onboard object detection models that maintain accuracy under noisy conditions while running efficiently on standard FPGA accelerators.
- •For embedded system developers: Develop spaceborne embedded vision systems combining CNN and transformer components without modifying existing FPGA accelerator architectures.
Authors
Adrien Dorise, Marjorie Bellizzi, Julia Cohen, Stéphane May
Abstract
Onboard object detection in Earth observation is constrained by limited computational resources and the absence of fully corrected imagery. While convolutional detectors are hardware-efficient, they often struggle to extract robust representations from raw and noisy data. Conversely, transformer-based models provide stronger global reasoning capabilities but remain difficult to deploy on FPGA accelerators due to quadratic attention complexity and non-compatible operations. We introduce TriCCOT, a tri-part architecture for robust and deployable onboard object detection. TriCCOT combines a convolutional region proposal network, a conformal prediction stage, and Aper-GATES, our hardware-friendly attention-based classifier. The region proposal network generates candidate bounding boxes, which are subsequently enlarged via conformal prediction, providing a distribution-free probabilistic coverage guarantee. The resulting crops are processed by Aper-GATES, which reformulates self-attention through convolutional projections, global channel statistics, and hardware-friendly gating operations, avoiding standard transformer operations that are poorly suited to CNN-oriented accelerators. Experiments on the DIOR and VDVRaw datasets demonstrate competitive detection performance and improved robustness to spatial blur and signal-dependent noise when compared to FPGA-compatible architectures. Finally, we report full deployment on a Xilinx Versal VCK190 FPGA without modifying the underlying DPU architecture, enabling unified CNN-Transformer inference for spaceborne embedded applications.