Stage-aware 6DoF pose tracking of industrial objects for augmented reality assembly guidance
Measurement Science and Technology · 13.03.2026 · 37 (10) · 106305
DOI: 10.1088/1361-6501/ae4e01 ↗Kurz gefasst
Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 17.03.2026. Verfasst von Kai Liu und 4 weiteren Autor:innen. Die Arbeit nennt 44 Literaturangaben.
Abstract (Originalsprache)
Accurate and real-time six-degrees-of-freedom pose measurement of texture-less objects is critical for augmented reality (AR)-assisted assembly in smart manufacturing and medical applications. While contour-based tracking methods have demonstrated strong potential for tracking such objects, their tracking accuracy and stability tend to deteriorate in industrial assembly environments due to persistent contour mismatches in static occlusion regions caused by already assembled parts. To address this issue, we propose a novel stage-aware pose tracking method for AR assembly guidance. By leveraging the known geometric relationships between parts, we identify and exclude statically occluded contour regions to generate stage-specific sparse viewpoint models during the offline phase. During online tracking, the corresponding viewpoint model is selected according to the current assembly stage, ensuring that contour correspondences are established only from not statically occluded contour points. In addition, we introduce a hybrid contour extraction technique that adaptively fuses edge and foreground probability gradients to enhance the accuracy of candidate points under challenging industrial conditions such as color ambiguity and cluttered backgrounds. To further overcome dynamic occlusions introduced by human operators or moving machinery, we design a contour-constrained loss function incorporating both gradient and distance cues for hybrid pose optimization. For evaluation, we present two semi-synthetic datasets comprising six industrial parts and three assembly assets at different assembly stages, respectively. Extensive experiments on the public RBOT, our semi-synthetic datasets and real-world scenes demonstrate that our method outperforms existing approaches in tracking accuracy and robustness, showing its strong potential for practical deployment in AR-guided industrial assembly.
Zitieren
APA
Liu, K., Chen, J., Chen, J., Wang, L., & Wang, Y. (2026). Stage-aware 6DoF pose tracking of industrial objects for augmented reality assembly guidance. Measurement Science and Technology, 37(10), 106305. https://doi.org/10.1088/1361-6501/ae4e01
Vancouver
Liu K, Chen J, Chen J, Wang L, Wang Y. Stage-aware 6DoF pose tracking of industrial objects for augmented reality assembly guidance. Measurement Science and Technology. 2026;37(10):106305. doi:10.1088/1361-6501/ae4e01
BibTeX
@article{Liu2026,
author = {Liu, Kai and Chen, Jing and Chen, Jixiang and Wang, Leshan and Wang, Yongtian},
title = {Stage-aware 6DoF pose tracking of industrial objects for augmented reality assembly guidance},
journal = {Measurement Science and Technology},
year = {2026},
volume = {37},
number = {10},
pages = {106305},
doi = {10.1088/1361-6501/ae4e01}
}
EUVIMED