EUVIMEDEUVIMED entdecken
Crossref

TRPL-VIO: improved real-time thermal–inertial odometry with robust line feature tracking

Wenzhe Fan, Linyang Li, Dongqing Zhao, Luguang Lai, Xiongqing Li

Measurement Science and Technology · 01.05.2026 · 37 (18) · 186304

DOI: 10.1088/1361-6501/ae5abd ↗

Kurz gefasst

Fachartikel in der Zeitschrift Measurement Science and Technology, online erschienen am 08.05.2026. Verfasst von Wenzhe Fan und 4 weiteren Autor:innen. Die Arbeit nennt 38 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Natural Science Foundation for Excellent Young Scholars of Henan Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

In challenging scenarios, such as low-light and weak texture conditions, infrared (IR) camera–based visual positioning technologies are widely adopted. Beyond traditional point features, the introduction of line features has significantly enhanced the performance of IR simultaneous localization and mapping (SLAM) systems. However, conventional line detection methods are highly sensitive to the noise levels inherent in IR images, and existing line matching approaches primarily rely on descriptors, resulting in poor real-time performance and limited adaptability to IR imagery. To address these limitations, this study proposes an improved real-time thermal–inertial odometry system with robust line feature tracking (TRPL-VIO). The approach leverages the superior ELSED algorithm for line feature extraction and introduces an IR line feature matching strategy that combines optical flow with descriptor matching. Furthermore, an improved line feature reprojection error model is developed by comprehensively utilizing point-to-line distances and intersegment angles, and a sliding window optimization framework is implemented to optimize inertial measurement unit pre-integration factors along with point and line feature factors. Extensive experiments on both public datasets and customized underground datasets demonstrate the effectiveness of the proposed method. The results showed that TRPL-VIO AZE current advanced monocular visual–inertial system (VINS). Besides, The performance of the proposed line feature extraction and tracking module is superior to those of PL-VINS and TPL-SLAM, and can meet the real-time requirements of thermal inertial odometry applications. The corresponding code will be open-sourced on GitHub ( https://github.com/alksfjhlkansf/TRPL-VIO ).

Förderung

  • National Natural Science Foundation of China
  • Natural Science Foundation for Excellent Young Scholars of Henan Province

Zitieren

APA

Fan, W., Li, L., Zhao, D., Lai, L., & Li, X. (2026). TRPL-VIO: improved real-time thermal–inertial odometry with robust line feature tracking. Measurement Science and Technology, 37(18), 186304. https://doi.org/10.1088/1361-6501/ae5abd

Vancouver

Fan W, Li L, Zhao D, Lai L, Li X. TRPL-VIO: improved real-time thermal–inertial odometry with robust line feature tracking. Measurement Science and Technology. 2026;37(18):186304. doi:10.1088/1361-6501/ae5abd

BibTeX
@article{Fan2026,
  author = {Fan, Wenzhe and Li, Linyang and Zhao, Dongqing and Lai, Luguang and Li, Xiongqing},
  title = {TRPL-VIO: improved real-time thermal–inertial odometry with robust line feature tracking},
  journal = {Measurement Science and Technology},
  year = {2026},
  volume = {37},
  number = {18},
  pages = {186304},
  doi = {10.1088/1361-6501/ae5abd}
}
Kostenlos starten →