This Unreal Engine project serves as the simulation environment and dataset generator for the paper:
NeRF-based simultaneous pose estimation and 3D reconstruction for non-cooperative space target
Aerospace Science and Technology, 2025.
DOI: 10.1016/j.ast.2025.111010
This project contains high-fidelity 3D models of space targets and Earth environments used to generate synthetic training and testing data for NeRF-based pose estimation and reconstruction algorithms.
- Unreal Engine 5.3
- Visual Studio 2022 (or compatible C++ compiler for UE5)
space_earth.umap: Main simulation scene with Earth background.deep_space.umap: Deep space environment.small_4.umap: Additional test scene.
- Satellites: Includes models for various satellites such as:
- MRO
- Proba-2
- Ven
- Smart Satellite 1A (SS1A)
- Earth: High-resolution 8k textures for Earth (Day, Night, Clouds, Normal, Specular) located in
/Content/earth/.
- Clone this repository.
- Open
s_claw.uprojectwith Unreal Engine 5.3. - Navigate to
Content/to open thedeep_spaceLevel. - Use the Sequencer
around_small_room4to play back the cinematic sequences or render out image data. - If you wish to test different satellite models, you can download them from Google Drive. After decompression, place it in the
/Content/directory.
If you use this dataset or project in your research, please cite the following paper:
@article{YourPaper2025,
title = {NeRF-based simultaneous pose estimation and 3D reconstruction for non-cooperative space target},
journal = {Aerospace Science and Technology},
year = {2025},
doi = {10.1016/j.ast.2025.111010},
url = {https://doi.org/10.1016/j.ast.2025.111010}
}This project is provided for research purposes. Please refer to the paper for more details on usage rights and restrictions.
We would like to thank Chuan Yan for preparing the 3D models of space targets used in this UE project.