Answers are grounded in this paper and cite their sources.
Accurate SAR-AIS association is critical for maritime monitoring but faces three key challenges: 1) The sampling frequency of AIS varies irregularly, making it difficult to predict the trajectory. 2) The available features of sailing and anchored targets are not uniform and the distribution of position errors is different among different motion states, making conventional non-discriminative methods less universally applicable in different scenes. 3) The false alarms and missed detections in target-dense region which seriously affect the optimal association, result many wrong and missed associations. To solve the above problems, firstly, a Dual stage AIS Encoding (DAE) module is constructed to extract the features of AIS and unify feature dimensions. Secondly, a dual Geometric Transformer module is proposed to simultaneously consider the features of sailing and anchored targets as well as target topology and deriving the association results via sinkhorn method. Thirdly, overlap loss is introduced to consider the feature differences between SAR and anchored AIS, and restrict the wrong and missed association. The experimental results show that based on the simulation data, it improves 11.5-20.7%, 5.5-14.1%, and 9.1-20.4% in precision, F1 score, and overall accuracy respectively, compared with the GNN, DS evidence theory, Fuzzy decision-making method and IT2FS; based on the real data, it improves 4.3-7.9%, 1.8-5.6%, and 3.5- 10.2%, respectively. The proposed method yields more accurate and reliable association results.
No takes yet. Share an insight, caveat, or question.
Gao et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: