Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 27 of the 27 references it cites were found in a registry or an index; 14 of the 23 sources its claims cite were checked for support.

Pose Estimation for Cross-Domain Non-Cooperative Spacecraft Based on Spatial-Aware Keypoints Regression

Open access
Wang, Zihao, Liu, Yunmeng, Zhang, E · 2024 · Aerospace

Abstract

Reliable pose estimation for non-cooperative spacecraft is a key technology for in-orbit service and active debris removal missions. Utilizing deep learning techniques for processing monocular camera images is effective and is a hotspot of current research. To reduce errors and improve model generalization, researchers often design multi-head loss functions or use generative models to achieve complex data augmentation, which makes the task complex and time-consuming. We propose a pyramid vision transformer spatial-aware keypoints regression network and a stereo-aware augmentation strategy to achieve robust prediction. Specifically, we primarily use the eight vertices of a cuboid satellite body as landmarks and the observable surfaces can be transformed by, respectively, using the pose labels. The experimental results on the SPEED+ dataset show that by using the existing EPNP algorithm and pseudo-label self-training method, we can achieve high-precision pose estimation for target cross-domains. Compared to other existing methods, our model and strategy are more straightforward. The entire process does not require the generation of new images, which significantly reduces the storage requirements and time costs. Combined with a Kalman filter, the robust and continuous output of the target position and attitude is verified by the SHIRT dataset. This work realizes deployment on mobile devices and provides strong technical support for the application of an automatic visual navigation system in orbit.

Overview

CoChat Summary

TL;DR

The paper proposes PVSAR, a PVTv2-based keypoint regression network with spatial-aware augmentation and online pseudo-label self-training for monocular cross-domain non-cooperative spacecraft pose estimation. On SPEED+, it reports S_pose 0.076 on lightbox and 0.112 on sunlamp, and validates filtered pose outputs on SHIRT.

Key findings

  • PVSAR achieves Lightbox S_t/S_R/S_pose of 0.018/0.057/0.076 and Sunlamp 0.023/0.089/0.112, compared with SPNv2 baseline at 0.025/0.097/0.122 and 0.027/0.170/0.197, respectively.
  • The paper reports a 37.7% improvement over SPNv2 and a 24.7% improvement over VPU in the lightbox domain, with a 30.6M-parameter model versus SPNv2’s 52.5M, EagerNet’s >88M, and VPU’s 190.1M.
  • Ablation shows stereo-aware augmentation improves cross-domain S_pose: Lightbox from 0.8041 with only normalization to 0.1431, and Sunlamp from 0.9183 to 0.5677.
  • On SHIRT lightbox, sequence 2 matched the SPEED+ score, while sequence 1 dropped to 0.175; a simple previous-frame filter when S_pose exceeds 2.0 improved sequence 1 to 0.135, and a Kalman-filter pipeline achieved steady-state rotation errors at degree level and translation errors at centimeter level, with average error less than 5° or 5 cm.

Methods

  • Pose estimation is reformulated as direct regression of 11 keypoints—eight cuboid body vertices and three antenna vertices—followed by MAGSAC PnP; the network uses a PVTv2-B1 backbone, five deconvolution layers, keypoint and logit features, and SAR-style spatial positional priors without heatmap labels.
  • Training uses 47,966 synthetic SPEED+ samples and 11,994 validation samples, input resized from 1920×1200 to 768×480, AdamW with initial learning rate 0.0001 and batch size 8 for 50 epochs; online self-training uses learning rate 0.00001, inlier thresholds 9, 8, and 7 for 10 epochs each, totaling 30 epochs.
  • Inference uses PnP confidence 0.99 and reprojection error 20.0; SHIRT validation applies an unscented Kalman filter and low-pass filter, with quaternion converted to r6d representation.

Limitations

  • The paper states that online pseudo-label self-training learns from labels containing errors, inevitably introduces noise, and makes model forgetting almost unavoidable; relaxing inlier constraints may improve average scores but can shift the defined 3D keypoint positions.
  • The paper notes keypoint ambiguity among the eight body vertices and three antenna vertices, where incorrect keypoint predictions can receive high confidence; worst samples arise from suboptimal lighting or exposure in lightbox and harsh lighting plus background noise in sunlamp, and rotation error is not normalized like translation error, so keypoint misidentification can flip the pose and maximize rotation error.
  • The paper states SHIRT sequence 1 dropped to 0.175 because the target was far away and blended into the background under suboptimal lighting; future work is needed to improve the tracking pose navigation filter and investigate a lightweight architecture. The text also makes evident that validation is limited to SPEED+ and SHIRT datasets rather than real in-orbit imagery.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3390/aerospace11110948

Cites

27 of 27 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Verification

Claims the paper makes

10 Valid · 14 Not valid · 5 Not checked · 51 Own results and interpretations

27 more claims

Create a free account to see all 80 claims and to verify the 1 reference still open.

See all claims — free account
  • Probably not a substantial claim. 30
  • It may state two separate claims as one. 27
  • Borderline: one of the tests for a substantial claim is unclear. 1
  • The quote only partly states this claim. 2

Create a free account to see these claims.

Citations

Cites

Show all 27 references

Research works printed but not linked: 1

  1. Wang, 2011, Reflection characteristics of on-orbit satellites based on BRDF, Opto-Electron. Eng., 38, 6

Metadata

Authors
Wang, Z., Liu, Y., & Zhang, E.
Affiliations
Chinese Academy of Sciences (CN)
Corresponding author
  • Liu, Yunmeng
  • Zhang, E
Venue
Aerospace
Publisher
MDPI AG
Published
2024-11-17
Volume
11
Issue
11
Pages
948
ISSN
2226-4310 (ISSN-L)
OpenAlex
W4404568728
Cited by
5
References
27
Citations by year
2026: 3 · 2025: 2
FWCI
0.86
Type
journal article
Language
en
Open access
Locations
  • Aerospace (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
Shanghai Pilot Program for Basic Research—Chinese Academy of Sciences, Shanghai Branch (JCYJ-SHFY-2022-004)
Topics
  • Physical Sciences › Engineering › Aerospace Engineering › Space Satellite Systems and Control
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
  • Physical Sciences › Engineering › Aerospace Engineering › Spacecraft Dynamics and Control
Keywords
monocular camera, Pyramid Vision Transformer, domain generalization, EPnP, active debris removal, visual navigation, Kalman filter, mobile device deployment

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-08, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-08 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Pose Estimation for Cross-Domain Non-Cooperative Spacecraft Based on Spatial-Aware Keypoints Regression. https://app.cochat.ai/papers/10.3390/aerospace11110948
The paper
Wang, Z., Liu, Y., & Zhang, E. (2024). Pose Estimation for Cross-Domain Non-Cooperative Spacecraft Based on Spatial-Aware Keypoints Regression. Aerospace. https://doi.org/10.3390/aerospace11110948

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.