.png){kind=icon}
.png){kind=icon}
Return
StellarGCN: A Star Identification Method Based on Graph Neural Network
.png){kind=icon}
PRE
AI
K
Y
H
Z
DOI：10.1109/taes.2026.3714379.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Star identification is a key component for ensuring the accuracy of autonomous spacecraft navigation. Classical methods often struggle under noisy conditions and complex backgrounds. In this article, we propose a star identification method based on graph neural networks (GNNs). Specifically, each star image is encoded as a graph representation by exploiting angular distances and spatial distribution among stars. Then, we introduce a GNN model termed stellarGCN, which combines graph convolution, graph pooling, and geometric information fusion to improve robustness under noisy conditions. Extensive experiments on synthetic star images demonstrate that the proposed method significantly outperforms representative baseline methods in both identification accuracy and stability. Our code is publicly available online.
Keywords:
Stars
Noise
Modeling
Distance measurement
Accuracy
Encoding
Graph neural networks
Algorithms
Convolution
Pixel
Journal
IF:
5.7
Papers:
651
Citations:
2.4W
Organization
Cited Papers