Vision-Based Detection of Uncooperative Targets and Components on Small Satellites

TL;DR

This work addresses space situational awareness for uncooperative targets by developing a distance-adaptive perception pipeline that switches between long-range LWIR-based detection using YOLOv8 (zero-shot and domain-specific) and short-range spacecraft-part segmentation via knowledge-distilled Fast-SCNN. It introduces a LWIR spacecraft dataset and demonstrates real-time long-range detection with AP50 ≈ 0.89 and AP75 ≈ 0.667 on a Jetson TX2 NX, enabling realtime tracking at ~60–90 ms per frame. For short-range segmentation, it distills features from a visual foundation model (Dino) into Fast-SCNN to achieve real-time, low-storage semantic segmentation on RGB imagery, with mixed results across custom and Adelaide datasets and clear benefits in data-constrained regimes. The results support onboard, edge-computing deployment for Edge Node-like missions, enabling robust long-range detection and efficient short-range segmentation to enhance spaceflight autonomy and safety.

Abstract

Space debris and inactive satellites pose a threat to the safety and integrity of operational spacecraft and motivate the need for space situational awareness techniques. These uncooperative targets create a challenging tracking and detection problem due to a lack of prior knowledge of their features, trajectories, or even existence. Recent advancements in computer vision models can be used to improve upon existing methods for tracking such uncooperative targets to make them more robust and reliable to the wide-ranging nature of the target. This paper introduces an autonomous detection model designed to identify and monitor these objects using learning and computer vision. The autonomous detection method aims to identify and accurately track the uncooperative targets in varied circumstances, including different camera spectral sensitivities, lighting, and backgrounds. Our method adapts to the relative distance between the observing spacecraft and the target, and different detection strategies are adjusted based on distance. At larger distances, we utilize You Only Look Once (YOLOv8), a multitask Convolutional Neural Network (CNN), for zero-shot and domain-specific single-shot real time detection of the target. At shorter distances, we use knowledge distillation to combine visual foundation models with a lightweight fast segmentation CNN (Fast-SCNN) to segment the spacecraft components with low storage requirements and fast inference times, and to enable weight updates from earth and possible onboard training. Lastly, we test our method on a custom dataset simulating the unique conditions encountered in space, as well as a publicly-available dataset.

Figures (8)

• Figure 1: Overview of the long-short detection architecture proposed for the Edge Node mission or other OOS tasks. EdgeNode We propose a long distance detection model on thermal images for detection at larger ranges into a segmentation model that identifies spacecraft parts for OOS on RGB images.
• Figure 2: Dataset samples with zoomed-in crops (red) used to develop models for long range spacecraft detection.
• Figure 3: Results of long range spacecraft detection using thermal imaging. To aid visualization, the spacecraft in ground truth samples are delineated with yellow bounding boxes.
• Figure 4: Example images from the spacecraft parts dataset generated in the spacecraft robotic simulator (top row) and the Adelaide dataset (bottom row).
• Figure 5: Results of short range spacecraft component detection using RGB imaging.