Robust Self-Supervised Monocular Visual Odometry via Lagged Temporal Trajectory Assisting Optimization
IEEE Transactions on Instrumentation and Measurement · 2026 · 75 · 5009114-5009114
DOI: 10.1109/tim.2026.3687358 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Instrumentation and Measurement, 2026. Verfasst von Yiyou Liang und 4 weiteren Autor:innen. Die Arbeit nennt 39 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Scientific and Technological Innovation Foundation of Foshan.
Abstract (Originalsprache)
Self-supervised visual odometry (VO) methods estimate camera ego-motion from unlabeled image sequences, reducing the need for costly labeled data. However, these methods often rely on the grayscale invariance assumption, which leads to performance degradation under significant brightness variations. Additionally, existing approaches frequently fail to fully leverage temporal information in image sequences. To address these limitations, we propose an appearance feature enhancement module to improve the representation of input images and introduce a nonlinear brightness alignment photometric loss to mitigate the impact of brightness inconsistencies. Furthermore, we design a lagged temporal trajectory optimization network based on a multitask learning strategy. This network effectively integrates temporal information and optimizes pose estimation results by sharing encoder weights with the pose estimation network, enhancing feature representation through joint training. Extensive experiments demonstrate the superior performance of our proposed method.
Förderung
- National Natural Science Foundation of China
- Scientific and Technological Innovation Foundation of Foshan
Zitieren
APA
Liang, Y., Zeng, H., Yi, F., Ren, Y., & Jiao, X. (2026). Robust Self-Supervised Monocular Visual Odometry via Lagged Temporal Trajectory Assisting Optimization. IEEE Transactions on Instrumentation and Measurement, 75, 5009114-5009114. https://doi.org/10.1109/tim.2026.3687358
Vancouver
Liang Y, Zeng H, Yi F, Ren Y, Jiao X. Robust Self-Supervised Monocular Visual Odometry via Lagged Temporal Trajectory Assisting Optimization. IEEE Transactions on Instrumentation and Measurement. 2026;75:5009114-5009114. doi:10.1109/tim.2026.3687358
BibTeX
@article{Liang2026,
author = {Liang, Yiyou and Zeng, Hui and Yi, Fangzhou and Ren, Yang and Jiao, Xikun},
title = {Robust Self-Supervised Monocular Visual Odometry via Lagged Temporal Trajectory Assisting Optimization},
journal = {IEEE Transactions on Instrumentation and Measurement},
year = {2026},
volume = {75},
pages = {5009114--5009114},
doi = {10.1109/tim.2026.3687358}
}
EUVIMED