Three-Dimensional vibration measurement of rotating machinery based on binocular vision and 3D bounding box
Measurement Science and Technology · 10.04.2026 · 37 (15) · 156008
DOI: 10.1088/1361-6501/ae5d60 ↗Kurz gefasst
Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 17.04.2026. Verfasst von Fameng Yang und 2 weiteren Autor:innen. Die Arbeit nennt 49 Literaturangaben. Gefördert von: Xingdian Talent Support Plan of Yunnan Province, National Natural Science Foundation of China.
Abstract (Originalsprache)
In the field of visual vibrometry, binocular vision has been employed to measure two-dimensional vibration displacements of rotating machinery; however, accurately capturing three-dimensional (3D) spatial vibrations remains a significant challenge. Addressing the technical bottlenecks associated with weakly textured surfaces and submillimeter-level 3D vibration measurements in regularly shaped rotating machinery, the study proposes an innovative measurement framework that integrates non-parallel binocular vision with deep learning techniques. Through high-precision camera calibration, both intrinsic and extrinsic parameters are accurately determined. To enhance object detection performance for rotating machinery, we innovatively modify the you only look once v11 deep learning network by incorporating a focal modulation block and an efficient up-convolution block. These enhancements enable precise extraction of multiple key points within detected bounding boxes, facilitating the construction of a structured seven-class keypoint matching system. By applying extremal constraints algorithms to establish 3D bounding box models, accurate centroid coordinate computation is achieved. The experiments were conducted by constructing a non-parallel binocular vision system on a rotor vibration test rig to capture vibration image data and calibration images of the rotor. Comparative analysis was performed against reference measurements obtained from eddy current displacement sensor, complemented by ablation studies and comprehensive evaluations involving multiple algorithms. The results demonstrate that the proposed method achieves mean root mean square errors of 0.025 mm across the X , Y , and Z axes, while reducing the average peak frequency tracking error to as low as 0.463 × 10 − 2 Hz. These findings fully confirm its high precision in 3D vibration measurement for objects with regular geometries, providing an innovative non-contact solution for micro-vibration monitoring in industrial-grade rotating machinery.
Förderung
- Xingdian Talent Support Plan of Yunnan Province
- National Natural Science Foundation of China
Zitieren
APA
Yang, F., Wang, S., & Zeng, X. (2026). Three-Dimensional vibration measurement of rotating machinery based on binocular vision and 3D bounding box. Measurement Science and Technology, 37(15), 156008. https://doi.org/10.1088/1361-6501/ae5d60
Vancouver
Yang F, Wang S, Zeng X. Three-Dimensional vibration measurement of rotating machinery based on binocular vision and 3D bounding box. Measurement Science and Technology. 2026;37(15):156008. doi:10.1088/1361-6501/ae5d60
BibTeX
@article{Yang2026,
author = {Yang, Fameng and Wang, Sen and Zeng, Xiangjun},
title = {Three-Dimensional vibration measurement of rotating machinery based on binocular vision and 3D bounding box},
journal = {Measurement Science and Technology},
year = {2026},
volume = {37},
number = {15},
pages = {156008},
doi = {10.1088/1361-6501/ae5d60}
}
EUVIMED