Target Intention Recognition Based on Multi-Type Information Fusion and DS Ensemble Classification in Aerial Combat

Aerial combat target intention recognition is a critical component of combat situational awareness, providing essential support for combat decision-making. Due to the robust learning capabilities of neural networks, neural network-based methods for target intention recognition have garnered significant research attention. However, existing methods often overlook the heterogeneity of information types, the latent information in future dynamic behaviors, and exhibit a lack of reliability in complex combat scenarios. To address these issues, this paper proposes a neural network model that integrates multi-type information fusion and Dempster-Shafer (DS) ensemble classification for aerial combat target intention recognition. The proposed model first performs classification based on data attributes and incorporates a prediction module, a neural network module, and an attention mechanism for the expansion, extraction, and fusion of feature information. Additionally, the model constructs an ensemble classifier grounded in ensemble learning principles and integrates the classification results using DS evidence theory to enhance reliability. The superiority and effectiveness of the proposed method are validated through comprehensive comparison and ablation experiments.