MFA-Net:一种面向复杂对抗环境的反舰导弹智能识别多模态自适应融合网络

打开文本图片集

中图分类号：TJ760 文献标志码：A DOI： 10.12305/j.issn.1001-506X.2026.03.16

Abstract：Aiming at the problems of blurred features， strong decoy deception， and insufficient robustness of traditionalrecognitionalgorithmsfacedbyanti-shipmisilestargetrecognitionincomplexadversarialenvironments，a multimodal adaptive fusionnetwork（MFA-Net）isproposed.The model employs parameter non-shared branches to extract heterogeneous features from radar，infrared，and electronic support measures，and achieves cros-modal adaptive fusion throughadual-dimensional channel-spaceatention mechanism.Byintroducinganadversarialtraining strategybasedonthe momentum iterative method，the modelenhances its intrinsic robustnesswithinaminimax optimization framework，thereby improving decision stability under jamming conditions.Anonlinear comprehensive recognition efectiveness index（CREI）is constructedbyintegratingthe Macro-F1score，anti-jammingrobustness，and inference timeliness.Experimentalresults demonstrate thatthe MFA-NetachievesaCREI of0.8531，significantly outperformingseveral comparative models. Sensitivityanalysis of jamming intensity furtervalidated the performance stabilityof the model under diferent levels of adversarial attacks.

Keywords：deep learming；adaptive fusion；dual-imensional attention；inteligent recognition；momentum iterative method

0 引言

现代海战已演变为体系与体系的激烈对抗，制海权的争夺愈发依赖于由侦察、指挥、打击、保障等节点构成的装备体系的整体作战效能[。（剩余26672字）