A ground-based dataset and a diffusion model for on-orbit low-light image enhancement

TL;DR

This work tackles on-orbit low-light image enhancement by addressing data scarcity through a ground-based Beidou satellite LLIE dataset collected with a 6-DoF robot in a collision-free, stratified sampling setup. It then presents a diffusion-model LLIE framework enhanced with fused-attention guidance to preserve structure while brightening dark regions. The method demonstrates superior performance over traditional and several deep-learning baselines and shows practical benefits for downstream tasks through improved feature richness. The findings offer a data-efficient path toward robust space-environment sensing and inform future multi-target and multi-task diffusion approaches in space imaging.

Abstract

On-orbit service is important for maintaining the sustainability of space environment. Space-based visible camera is an economical and lightweight sensor for situation awareness during on-orbit service. However, it can be easily affected by the low illumination environment. Recently, deep learning has achieved remarkable success in image enhancement of natural images, but seldom applied in space due to the data bottleneck. In this article, we first propose a dataset of the Beidou Navigation Satellite for on-orbit low-light image enhancement (LLIE). In the automatic data collection scheme, we focus on reducing domain gap and improving the diversity of the dataset. we collect hardware in-the-loop images based on a robotic simulation testbed imitating space lighting conditions. To evenly sample poses of different orientation and distance without collision, a collision-free working space and pose stratified sampling is proposed. Afterwards, a novel diffusion model is proposed. To enhance the image contrast without over-exposure and blurring details, we design a fused attention to highlight the structure and dark region. Finally, we compare our method with previous methods using our dataset, which indicates that our method has a better capacity in on-orbit LLIE.

Figures (14)

• Figure 1: The illustration of on-orbit service and low-loght images. (a) demonstrates service star refueling the target star. (b) and (c) are the real images of low-light and normal-light captured by a visible camera during PRISMA mission49.
• Figure 2: The layout of the ground-test platform. t is composed of a 6 DoF robot carrying a satellite model spinning and a calibrated camera.
• Figure 3: The virtual environment in Pybullet. On the left is the visual rendered model, on the right is the collision mesh of the model.
• Figure 4: The $x-z$ and $y-z$ sectional view of the collision-free working space.
• Figure 5: The comparison of radius and elevation of different sampling times. The first column is the radius of the camera in the satellite spherical coordinates after different times of sampling. The second column is the elevation of the camera in the satellite spherical coordinates after different times of sampling.