EUVIMEDEUVIMED entdecken
Crossref

Rail3DNet: A Monocular Vision-Based 3-D Rail-Track Detection Network Using Rail-IoU Loss

Songyue Yang, Hongbo Li, Zhangyu Wang, Peng Chen, Guizhen Yu, Junjie Zhang

IEEE Transactions on Intelligent Transportation Systems · 2026-10 · 27 (10) · 12911-12924

DOI: 10.1109/tits.2026.3713896 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Intelligent Transportation Systems, 2026-10. Verfasst von Songyue Yang und 5 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 56 Literaturangaben. Gefördert von: National Natural Science Foundation of China, China Postdoctoral Science Foundation, National Key Technologies Research and Development Program of China.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Three-dimensional Rail-track detection is crucial for autonomous driving trains. Disadvantageously, existing methods primarily detect the rail-tracks on the image plane, making it difficult to obtain the three-dimensional coordinates of the rail-tracks. Therefore, this study introduces the first monocular vision-based rail-track detection network (Rail3DNet). An information interaction module was proposed to fully utilize the prior information on the rail-track. Besides, a feature reconstruction method was introduced to reconstruct the extracted features and incorporate three-dimensional coordinate information, thereby enriching the feature information content. Furthermore, a three-dimensional rail intersection over union loss function was designed to align the rail-tracks closer to the ground truth. The experimental results revealed that the proposed method had achieved an mF1 score of 73.58%, which was characteristic of state-of-the-art performance. The effectiveness of the proposed method was also validated on three different rail transit lines. The proposed rail-track detection network can offer accurate train clearances and increased safety in the navigation of autonomous trains.

Förderung

  • National Natural Science Foundation of China
  • China Postdoctoral Science Foundation
  • National Key Technologies Research and Development Program of China
  • Beijing Natural Science Foundation
  • Advanced Laser Technology Laboratory of Anhui Province

Zitieren

APA

Yang, S., Li, H., Wang, Z., Chen, P., Yu, G., & Zhang, J. (2026). Rail3DNet: A Monocular Vision-Based 3-D Rail-Track Detection Network Using Rail-IoU Loss. IEEE Transactions on Intelligent Transportation Systems, 27(10), 12911-12924. https://doi.org/10.1109/tits.2026.3713896

Vancouver

Yang S, Li H, Wang Z, Chen P, Yu G, Zhang J. Rail3DNet: A Monocular Vision-Based 3-D Rail-Track Detection Network Using Rail-IoU Loss. IEEE Transactions on Intelligent Transportation Systems. 2026;27(10):12911-12924. doi:10.1109/tits.2026.3713896

BibTeX
@article{Yang2026,
  author = {Yang, Songyue and Li, Hongbo and Wang, Zhangyu and Chen, Peng and Yu, Guizhen and Zhang, Junjie},
  title = {Rail3DNet: A Monocular Vision-Based 3-D Rail-Track Detection Network Using Rail-IoU Loss},
  journal = {IEEE Transactions on Intelligent Transportation Systems},
  year = {2026},
  volume = {27},
  number = {10},
  pages = {12911--12924},
  doi = {10.1109/tits.2026.3713896}
}
Kostenlos starten →