IRT Saint Exupery is proud to publish an open dataset dedicated to vision based navigation for space rendezvous !
Space rendezvous, the capability for two space satellites to make contact, is crucial for the European space industry, enabling advancements in orbit debris management, human exploration missions, and space services like repairs and transport. At IRT Saint Exupery, we’re developing key technologies to make space rendezvous autonomous and cost-effective.
Vision-based pose estimation (i.e. estimation of the position and orientation of the target vehicle) using deep learning offers a promising cost effective and versatile solution for relative satellite navigation. However, reliance on large simulated datasets for training necessitates bridging the gap between synthetic and real-world data.
To tackle this, a dataset of 120,000 images has been created using an novel approach to generate synthetic space scene images, ensuring fine statistical balancing for training and evaluation. Our synthetic images incorporate a physically based camera model for realism and include additional information like masks, distance maps, celestial body positions, and precise camera parameters for enhanced utility in training and testing.
The results of the work have been published in the 2024 AI4Space workshop
The full dataset is available for download:
This work has been accomplished in the frame of the RAPTOR project on the behalf of Thales Alenia Space.