Qiming Liu

Dense 3D reconstruction and accurate pose measurement of non-cooperative satellites are essential for on-orbit servicing missions. Multi-view stereo (MVS) is the preferred approach for high-precision dense reconstruction from monocular imagery, yet current methods struggle with the speed–accuracy trade-off and reconstruction degradation under complex on-orbit conditions, while keypoint-based pose measurement is further constrained by limited training data and heavy reliance on manual annotations. To address these challenges, we present a monocular vision-based method coupling dense 3D reconstruction with pose estimation. We develop MambaMVS, an end-to-end MVS network featuring a Mamba Cost Volume (MCV) Module for efficient cost volume regularization via the linear complexity of structured state space models, and an improved focused linear attention module for robust cross-view feature matching. The reconstructed point cloud provides 3D keypoint coordinates for pose measurement, where a few-shot scheme employing 3D Gaussian Splatting-based view augmentation and reprojection-driven self-labeling enables training with minimal manual annotations. During online inference, detected 2D keypoints are combined with offline-reconstructed 3D coordinates via the Perspective-n-Point algorithm for fast solving of the 6-DOF pose. Experiments on the Customized Space Target dataset demonstrate state-of-the-art reconstruction quality with 1.5× faster inference than current leading methods. Physical satellite model experiments achieve rotation ME within 0.2° and translation ME below 0.01 m, while robustness is validated across varying illumination, reflective surfaces, motion blur, and occlusion. The proposed reconstruction-guided measurement paradigm offers a practical pathway toward autonomous on-orbit perception and in situ structural assessment.

Far-field vision-based pose measurement is a crucial technology for applications such as high-resolution Earth observation and space security early warning. However, owing to the perspective imaging model of long-range optical systems, conventional vision-based pose measurement methods are highly susceptible to image noise and pose parameter coupling, leading to significant estimation deviations. Consequently, these methods fail to meet the rigorous requirements for the accurate measurement and intelligent perception of object poses in far-field scenarios, particularly when the object distance significantly exceeds the focal length. To address these challenges, this paper presents a Perspective-n-Point (PnP) preprocessing and post-optimization method for far-field pose measurement based on weighted central normalization. First, the Robust PnP (RPnP) algorithm is employed to obtain an initial pose for the far-field object, and an objective function is formulated by minimizing the reprojection error of the image feature points. Second, central normalization is applied to the Jacobian matrix of the pose parameters, and the information matrix is weighted according to the localization uncertainty of the image feature points. Finally, a weighted nonlinear optimization is executed to obtain refined pose parameters. Under the tested conditions, this approach can reduce the sensitivity of the pose parameters to image noise, minimizes the coupling among extrinsic parameters, and reduces the tendency of noise-driven pose-update excursions. The proposed method is evaluated through simulations and scaled physical relative-comparison experiments, supporting its potential for numerically stable vision-based pose measurement of far-field objects in aerospace and related domains. Noise-and-turbulence simulations demonstrate the pose-refinement benefit of CS and improved rotation estimation with a known spatial covariance model.