LiDAR-FMC: Accurate and Robust Human Capture From Point-Cloud Video
IEEE Transactions on Circuits and Systems for Video Technology · 2026-07 · 36 (7) · 9808-9820
DOI: 10.1109/tcsvt.2026.3670967 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Circuits and Systems for Video Technology, 2026-07. Verfasst von Bohao Fan und 3 weiteren Autor:innen. Die Arbeit nennt 42 Literaturangaben. Gefördert von: National Natural Science Foundation of China.
Abstract (Originalsprache)
This paper proposes a framework for online human motion capture based on point-cloud videos, aimed at achieving accurate human mesh reconstruction and smooth motion capture. Due to the inherently unordered and sparse nature of point cloud data, most existing human motion capture algorithms struggle to simultaneously achieve high accuracy and temporal consistency. To extract dynamic information from point cloud video sequences explicitly, this paper proposes a feature extraction network based on scene flow estimation. By leveraging consecutive two-frame point cloud inputs, the network explicitly captures temporal positional changes of local point sets. Utilizing these motion cues, our proposed scene-Flow-based Motion Capture network (termed LiDAR-FMC) simultaneously accomplishes accurate human mesh reconstruction and robust motion capture. It also achieves a throughput of over 17 fps when processing point cloud videos. Evaluations on multiple public datasets demonstrate that the proposed approach not only attains state-of-the-art performance under standard settings, but also consistently delivers accurate capture results for arbitrarily long video inputs.
Förderung
- National Natural Science Foundation of China
Zitieren
APA
Fan, B., Zheng, W., Feng, J., & Zhou, J. (2026). LiDAR-FMC: Accurate and Robust Human Capture From Point-Cloud Video. IEEE Transactions on Circuits and Systems for Video Technology, 36(7), 9808-9820. https://doi.org/10.1109/tcsvt.2026.3670967
Vancouver
Fan B, Zheng W, Feng J, Zhou J. LiDAR-FMC: Accurate and Robust Human Capture From Point-Cloud Video. IEEE Transactions on Circuits and Systems for Video Technology. 2026;36(7):9808-9820. doi:10.1109/tcsvt.2026.3670967
BibTeX
@article{Fan2026,
author = {Fan, Bohao and Zheng, Wenzhao and Feng, Jianjiang and Zhou, Jie},
title = {LiDAR-FMC: Accurate and Robust Human Capture From Point-Cloud Video},
journal = {IEEE Transactions on Circuits and Systems for Video Technology},
year = {2026},
volume = {36},
number = {7},
pages = {9808--9820},
doi = {10.1109/tcsvt.2026.3670967}
}
EUVIMED