Answers are grounded in this paper and cite their sources.
To address the issue of feature point localization errors in the automated assembly of high-voltage transmission tower cylinders, this paper proposes a relative pose estimation method based on outlier detection. The method first employs the PNP algorithm to compute an initial estimate of the relative pose between the camera and the tower cylinder. Subsequently, through reprojection error analysis, Z-score criteria are applied to detect and eliminate anomalous matching points, thereby constructing a highly reliable matching point set. Finally, PNP is re-solved based on the optimized point set to obtain precise relative pose estimation results. Experimental results demonstrate that this method effectively suppresses random errors caused by misdetected feature points, significantly improving the accuracy and stability of pose estimation. The mean angular error of the rotation matrix estimation around the X/Y axes does not exceed 3.07°, while the mean translation vector estimation error is below 3.10 mm. This research contributes to ensuring the precision and stability of automated assembly for high-voltage transmission tower cylinders.
No takes yet. Share an insight, caveat, or question.
Xia et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: