Spacecraft pose estimation using computer vision has garnered increasing attention in research areas such as automation system theory, control theory, sensors and instruments, robot technology, and automation software. Confronted with the extreme environment of space, existing spacecraft pose estimation methods are predominantly multi-stage networks with complex operations. In this study, we propose an approach for spacecraft homography pose estimation with a single-stage deep convolutional neural network for the first time. We formulated a homomorphic geometric constraint equation for spacecraft with planar features. Additionally, we employed a single-stage 2D keypoint regression network to obtain homography 2D keypoint coordinates for spacecraft. After decomposition to obtain the rough spacecraft pose based on the homography matrix constructed according to the geometric constraint equation, a loss function based on pixel errors was employed to refine the spacecraft pose. We conducted extensive experiments using widely used spacecraft pose estimation datasets and compared our method with state-of-the-art techniques in the field to demonstrate its effectiveness.
27 December 2023 in IEEE Transactions on Industrial Informatics
Linear servomotors (LSM) are widely used in industries. A LSM requires a mover positioning tool for precise control. A Vision-based digital image grating (DIG) measurement for LSM mover positioning is proposed in this study with advantages of simple system, low cost, and good robustness to environments. It estimates the LSM's rough position according to four mark-lines and obtains accurate LSM's positions by a fast least-square (FLS) image-matching algorithm. As a result, it can measure the LSM's position accurately in a wide range. The FLS proposed here represents the least-square cost function by the transformation parameters explicitly, and then calculates the parameters directly. Therefore, the image resampling process, which is inevitable for the existing least-square methods is not necessary. Simulations prove that the FLS is approximately 4.4 times faster than the classic inverse compositional Gauss–Newton (IC-GN) algorithm, and that the FLS is more accurate and robust than the IC-GN. Experiments show that the proposed DIG method can measure the LSM mover positions with an accuracy of 0.75 μ m mean absolute error.
30 July 2024 in IEEE Transactions on Industrial Informatics
Spatial target pose estimation with deep learning has garnered increasing attention in recent years. However, the existing methods in this field suffer from poor generalization. In this study, we propose a robust and reliable pose estimation method for spatial targets. The method aims to achieve keypoints filtrating. It involves a detection network tasked with identifying the target area, while the subsequent stage employs a classification network to regress keypoints from the detected target area. To improve the accuracy of pose estimation, we leverage spatial target geometric constraints to formulate 2-D–3-D keypoints equations for an initial pose. Then, we create a nonlinear optimization equation based on the confidence of 2-D keypoints and accomplish nonlinear refinement. We conduct extensive experiments on commonly used datasets and demonstrate the effectiveness of the proposed method. Furthermore, thanks to the effectiveness of keypoints filtrating and nonlinear refinement, the proposed method is robust with challenging scenarios and domain bias.
18 September 2024 in IEEE/ASME Transactions on Mechatronics
Uncooperative spacecraft pose estimation with neural networks is receiving increasing attention. However, the existing methods have poor generalization. In this research, we propose a pose estimation with normalized segmentation coordinate space for uncooperative spacecraft. We roughly calculate base points coordinates and equivalent diameter of uncooperative spacecraft through multiview rendezvous. We leverage base points coordinates and equivalent diameter to construct a 3D bounding box. Vertices coordinates of 3D bounding box need to be adjusted so that the box can segment the torso of uncooperative spacecraft proximately. We leverage convolutional neural network to detect spacecraft and obtain regions of interest in image. In addition, projection points of the box vertices in image can be regressed through classifier, while uncooperative spacecraft refined pose is obtained by inputting vertices projection information into pose estimation optimizer. We conduct extensive experiments on a commonly used datasets and demonstrated the superior performance of the proposed method. Furthermore, the proposed method is robust with challenging scenarios and domain bias. In addition, we design a hardware system for uncooperative spacecraft pose estimation to validate the effectiveness of our proposed method.
1 September 2024 in Journal of Aerospace Engineering
Research on computer vision-based unmarked satellite pose estimation is gaining increasing attention in fields such as intelligent control, industrial communication, and automation technology. A novel unmarked satellite pose estimation method based on semi-supervised domain adaptation and information augmentation is proposed. Efficient pseudo- label generation is achieved by utilizing the homography properties of satellites. Additionally, data enhancement is performed using limited target datasets, sparse satellite feature information, and image preprocessing. The augmented target dataset undergoes filtering and semi-supervised training through iterative merging of source and target datasets until optimization of the loss function related to pixel differences is accomplished. Finally, geometric constraint equations are constructed using the 2D keypoints and the 3D skeleton model of the satellite, with these equations being solved to obtain the satellite’s pose. Extensive experiments are conducted on common satellite pose estimation datasets, and the effectiveness of the proposed method is demonstrated.