Abstract:

To address the core issues of poor adaptability of static fusion strategies in existing recognition models, as well as the simplistic evaluation system and its disconnection from dynamic confrontation requirements, a practical four-dimensional evaluation system encompassing "recognition accuracy, anti-jamming stability, decision timeliness, and modal complementarity" was constructed, and an operational effectiveness composite index (OECI) capable of dynamically adapting to tactical scenarios was proposed. A multimodal dynamic attention fusion network (MDA-Net) for anti-ship missiles in complex confrontation environments was designed. Through heterogeneous feature decoupling, dynamic weighting of cross-modal attention, and a hierarchical gating decision mechanism, the autonomous evaluation and adaptive fusion of multi-source information reliability under jamming environments were realized. Simulation experiments indicate that under simulated high-intensity confrontation and modal missing scenarios, the OECI of MDA-Net significantly outperforms the baseline models of traditional multi-layer perceptron and convolutional neural networks, exhibiting excellent comprehensive recognition performance, anti-jamming resilience, and environmental adaptability. The constructed complete paradigm of "evaluation system, recognition model, and tactical adaptation" provides a systematic theoretical method and technical support for the design and effectiveness evaluation of intelligent recognition systems for anti-ship missiles.

Key words: anti-ship missile, autonomous recognition, multimodal fusion, feature engineering, effectiveness evaluation, intelligent combat