EUVIMEDEUVIMED entdecken
Crossref

Pose Estimation and Neural Implicit Reconstruction Toward Noncooperative Spacecraft Without Offline Prior Information

Bing Han, Chenxi Wang, Xinyu Zhang, Zhibin Zhao, Zhi Zhai, Jinxin Liu, Naijin Liu, Xuefeng Chen

IEEE Transactions on Aerospace and Electronic Systems · 2025-04 · 61 (2) · 2612-2630

DOI: 10.1109/taes.2024.3479199 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Aerospace and Electronic Systems, 2025-04. Verfasst von Bing Han und 7 weiteren Autor:innen. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 54 Literaturangaben. Gefördert von: National Natural Science Foundation of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Noncooperative spacecraft pose estimation plays a crucial role in on-orbit servicing. However, existing pose estimation methods often assume CAD models of target objects as prior information, used for offline training or online template matching. This limits the generalization of pose estimation methods. To explore a generic solution, this work proposes a pose estimation method for unknown spacecraft. Our method is not only independent of prior models or image priors of the target but also synchronously outputs pose parameters and aligned target texture models. Specifically, we employ three modules in parallel: pose tracking, neural object reconstruction, and target reference frame (TRF) estimation. First, leveraging the knowledge of temporal data, we optimize the pose graph to provide stable tracking performance. Then, we use neural implicit representation to reconstruct the target texture model, with pose parameters jointly optimized during the reconstruction process. Finally, we propose TRFE-Net for online estimation of the TRF. The obtained TRF is used to correct the sensor reference frame, transforming the pose tracking and reconstruction problem from scene-centric to spacecraft-centric. In addition, the PEU dataset was constructed specifically for pose estimation of unknown spacecraft. Comprehensive experiments show that although the proposed method reduces the need for prior information, it still achieves good performance across multiple objects and effectively handles large-scale motions, specular highlights, thin structures, and symmetric structures.

Förderung

  • National Natural Science Foundation of China

Zitieren

APA

Han, B., Wang, C., Zhang, X., Zhao, Z., Zhai, Z., Liu, J., Liu, N., & Chen, X. (2025). Pose Estimation and Neural Implicit Reconstruction Toward Noncooperative Spacecraft Without Offline Prior Information. IEEE Transactions on Aerospace and Electronic Systems, 61(2), 2612-2630. https://doi.org/10.1109/taes.2024.3479199

Vancouver

Han B, Wang C, Zhang X, Zhao Z, Zhai Z, Liu J, et al. Pose Estimation and Neural Implicit Reconstruction Toward Noncooperative Spacecraft Without Offline Prior Information. IEEE Transactions on Aerospace and Electronic Systems. 2025;61(2):2612-2630. doi:10.1109/taes.2024.3479199

BibTeX
@article{Han2025,
  author = {Han, Bing and Wang, Chenxi and Zhang, Xinyu and Zhao, Zhibin and Zhai, Zhi and Liu, Jinxin and Liu, Naijin and Chen, Xuefeng},
  title = {Pose Estimation and Neural Implicit Reconstruction Toward Noncooperative Spacecraft Without Offline Prior Information},
  journal = {IEEE Transactions on Aerospace and Electronic Systems},
  year = {2025},
  volume = {61},
  number = {2},
  pages = {2612--2630},
  doi = {10.1109/taes.2024.3479199}
}
Kostenlos starten →