LM-TTR: Lightweight Monocular Truck-Type Recognition for Source-Based Overload Control
Intelligent Computing · 01.01.2026 · 5
DOI: 10.34133/icomputing.0313 ↗Kurz gefasst
Fachartikel in der Zeitschrift Intelligent Computing, online erschienen am 19.08.2026. Verfasst von Ming Li und 7 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben. Gefördert von: Education and Scientific Research Project for Young and Middle-aged Teachers of Fujian Province, Foundation for Cultivated Young Talents of Fujian Province.
Abstract (Originalsprache)
Truck overloading compromises infrastructure and road safety. Since weight limits vary with axle configurations, accurate truck-type recognition is critical for intelligent source-based overload control systems. However, existing methods usually rely on full-vehicle global views or complex image stitching, which may lead to insufficient recognition accuracy, high deployment cost, or invasive road modification. To address these issues, we propose lightweight monocular truck type recognition (LM-TTR), which infers truck types from visible wheel-hub observations rather than full-vehicle appearance. Specifically, we develop RetinaHub, a lightweight wheel-hub detection network, to locate wheel hubs and recognize their categories from monocular video streams. We then introduce a spatial-prior-constrained neighborhood statistics mechanism to associate and aggregate the detected wheel hubs across frames, thereby deriving stable axle-level structural features, including axle displacement, axle-spacing distribution, and wheel-hub category distribution. Finally, these axle-level features are converted into a standardized code and matched with the standard truck-type library for truck-type recognition. Experimental results show that LM-TTR reduces processing time by more than 92% compared with existing algorithms and can run in real time on an embedded device with 0.5 teraflops computing capability. The proposed method achieves approximately 98% recognition accuracy for common truck types and demonstrates strong robustness under different illumination conditions.
Förderung
- Education and Scientific Research Project for Young and Middle-aged Teachers of Fujian Province
- Foundation for Cultivated Young Talents of Fujian Province
Zitieren
APA
Li, M., Ye, H., Xu, F., Luo, H., Guo, X., Yang, X., Zhang, J., & Yang, W. (2026). LM-TTR: Lightweight Monocular Truck-Type Recognition for Source-Based Overload Control. Intelligent Computing, 5. https://doi.org/10.34133/icomputing.0313
Vancouver
Li M, Ye H, Xu F, Luo H, Guo X, Yang X, et al. LM-TTR: Lightweight Monocular Truck-Type Recognition for Source-Based Overload Control. Intelligent Computing. 2026;5. doi:10.34133/icomputing.0313
BibTeX
@article{Li2026,
author = {Li, Ming and Ye, Huaisheng and Xu, Furong and Luo, Haibo and Guo, Xianwei and Yang, Xu and Zhang, Ji and Yang, Wencheng},
title = {LM-TTR: Lightweight Monocular Truck-Type Recognition for Source-Based Overload Control},
journal = {Intelligent Computing},
year = {2026},
volume = {5},
doi = {10.34133/icomputing.0313}
}
EUVIMED