EUVIMEDEUVIMED entdecken
Crossref

RwoTracker: wheel–rail lateral displacement detection algorithm based on point tracking

Xin Wang, Qingqiang Tan, Jiameng Duan, Jianyang Liu

Measurement Science and Technology · 22.05.2026 · 37 (21) · 215103

DOI: 10.1088/1361-6501/ae63ec ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 26.05.2026. Verfasst von Xin Wang und 3 weiteren Autor:innen. Die Arbeit nennt 33 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Effective wheel–rail contact is fundamental to the safe operation of railway vehicles. To address the susceptibility of conventional vision-based methods to motion blur and their lack of temporal consistency, this study proposes a novel dynamic trajectory tracking and measurement paradigm based on an enhanced CoTracker3 framework. First, a lightweight initial point selection network, RWPointNet, built upon MobileNetV4, is developed to enable targeted sampling and accurate initialization of key measurement points within the wheel–rail contact region. Second, a multi-scale four-dimensional correlation modeling strategy is introduced to capture pixel-level motion correspondences across both spatial and temporal domains. In addition, an as-rigid-as-possible constraint is incorporated as a physical prior to preserve the geometric consistency of the wheel–rail structure under high-speed motion, thereby effectively suppressing trajectory drift. Experimental results on a self-constructed dataset demonstrate an accuracy of 99.9% under an 8 pixel threshold, with a mean threshold accuracy of 84.8%, alongside substantial reductions in both mean absolute error (MAE) and root mean squared error. MAE is 1.31 pixels, corresponding to a physical displacement of 1.02 mm. Furthermore, the proposed method achieves an inference speed of 62 FPS on the Jetson Orin Nano edge computing platform. Overall, this work achieves a favorable balance between robustness and real-time performance, offering a promising new approach for wheel–rail displacement monitoring.

Zitieren

APA

Wang, X., Tan, Q., Duan, J., & Liu, J. (2026). RwoTracker: wheel–rail lateral displacement detection algorithm based on point tracking. Measurement Science and Technology, 37(21), 215103. https://doi.org/10.1088/1361-6501/ae63ec

Vancouver

Wang X, Tan Q, Duan J, Liu J. RwoTracker: wheel–rail lateral displacement detection algorithm based on point tracking. Measurement Science and Technology. 2026;37(21):215103. doi:10.1088/1361-6501/ae63ec

BibTeX
@article{Wang2026,
  author = {Wang, Xin and Tan, Qingqiang and Duan, Jiameng and Liu, Jianyang},
  title = {RwoTracker: wheel–rail lateral displacement detection algorithm based on point tracking},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {21},
  pages = {215103},
  doi = {10.1088/1361-6501/ae63ec}
}
Kostenlos starten →