High-Precision Multi-Object Tracking in Satellite Videos via Pixel-Wise Adaptive Feature Enhancement
Sensors · 09.10.2024 · 24 (19) · 6489
DOI: 10.3390/s24196489 ↗Kurz gefasst
Fachartikel in der Zeitschrift Sensors, online erschienen am 09.10.2024. Verfasst von Gang Wan und 7 weiteren Autor:innen. Beteiligte Einrichtung: School of Space Information, Space Engineering University, Beijing 101407, China. Die Arbeit wurde bisher 9-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Internal Parenting Program, State Key Laboratory of Geo-information Engineering.
Abstract (Originalsprache)
In this paper, we focus on the multi-target tracking (MOT) task in satellite videos. To achieve efficient and accurate tracking, we propose a transformer-distillation-based end-to-end joint detection and tracking (JDT) method. Specifically, (1) considering that targets in satellite videos usually have small scales and are shot from a bird’s-eye view, we propose a pixel-wise transformer-based feature distillation module through which useful object representations are learned via pixel-wise distillation using a strong teacher detection network; (2) targets in satellite videos, such as airplanes, ships, and vehicles, usually have similar appearances, so we propose a temperature-controllable key feature learning objective function, and by highlighting the learning of similar features during distilling, the tracking accuracy for such objects can be further improved; (3) we propose a method that is based on an end-to-end network but simultaneously learns from a highly precise teacher network and tracking head during training so that the tracking accuracy of the end-to-end network can be improved via distillation without compromising efficiency. The experimental results on three recently released publicly available datasets demonstrated the superior performance of the proposed method for satellite videos. The proposed method achieved over 90% overall tracking performance on the AIR-MOT dataset.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Space Information, Space Engineering University, Beijing 101407, China.
- State Key Laboratory of Geo-Information Engineering, Xi'an 710054, China.
ORCID-Profile: Zhijuan Su, Yitian Wu, Ningbo Guo
Förderung
- Internal Parenting Program
- State Key Laboratory of Geo-information Engineering
Zitieren
APA
Wan, G., Su, Z., Wu, Y., Guo, N., Cong, D., Wei, Z., Liu, W., & Wang, G. (2024). High-Precision Multi-Object Tracking in Satellite Videos via Pixel-Wise Adaptive Feature Enhancement. Sensors, 24(19), 6489. https://doi.org/10.3390/s24196489
Vancouver
Wan G, Su Z, Wu Y, Guo N, Cong D, Wei Z, et al. High-Precision Multi-Object Tracking in Satellite Videos via Pixel-Wise Adaptive Feature Enhancement. Sensors. 2024;24(19):6489. doi:10.3390/s24196489
BibTeX
@article{Wan2024,
author = {Wan, Gang and Su, Zhijuan and Wu, Yitian and Guo, Ningbo and Cong, Dianwei and Wei, Zhanji and Liu, Wei and Wang, Guoping},
title = {High-Precision Multi-Object Tracking in Satellite Videos via Pixel-Wise Adaptive Feature Enhancement},
journal = {Sensors},
year = {2024},
volume = {24},
number = {19},
pages = {6489},
doi = {10.3390/s24196489}
}
EUVIMED