基于时序知识图谱的空中目标行为意图识别方法OA

Behavioral intention recognition of aerial targets is a key component in situational awareness and command decision-making for modern air combat.To address the critical issues commonly encountered in existing deep learning models for behavior intention recognition-such as insufficient feature extraction,high misclassification rates,and inadequate interpretability-this paper proposes a behavior intention recog-nition method based on temporal knowledge graphs.First,domain knowledge is integrated with simulation data(BaseData)to construct a temporal knowledge graph for aerial combat(TLDATA)that captures temporal dependencies.Second,a Copy-Generation mechanism is introduced to enable precise transfer and generation of behavioral relational information within the graph,combined with a sliding-window historical memory module to effectively capture and leverage the dynamic evolution patterns of behaviors.Finally,a weighted fusion strategy is employed to integrate the model outputs,thereby enhancing recognition per-formance and robustness.Experimental results show that the proposed model outperforms baseline meth-ods on both the TLDATA and BaseData datasets,fully demonstrating its effectiveness and superiority.This work provides valuable insights and a useful reference for behavior intention recognition of aerial tar-gets in scenarios characterized by limited samples,low entropy,and high density in air combat environ-ments.