.png){kind=icon}
Return
SatSORT: Online multi-spacecraft visual tracking algorithm
.png){kind=icon}
PRE
AI
DOI：10.1016/j.actaastro.2025.02.027.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The deployment of large-scale constellations has led to a sharp increase in the number of spacecraft in orbit. avoid collision risks and overcome the limitations of traditional single spacecraft tracking methods, this introduces a multi-spacecraft visual tracking method called SatSORT tailored to the specific requirements aerospace field. Based on the DeepSORT framework, the proposed method eliminates the appearance descriptor and introduces a matching strategy tailored for the identical spacecrafts. Furthermore, the data association been improved to address the situations where targets are missed or occluded by each other. Enhancements have also been made to the Kalman filter to accommodate the varied visual motions of spacecraft and instances of missed detections. Extensive experimental results demonstrate that the new algorithm improves tracking performance impressively, reduces data association time by an order of magnitude, and maintains robust tracking capabilities under conditions of missed detections and occlusions.
Keywords:
Computer vision
Multi-object tracking
Satellite constellation
Occluded Object
Data association
Journal
IF:
3.4
Papers:
1.1W
Citations:
2.1W
Organization
Cited Papers
No cited papers available