基于时空感知双流融合网络的干扰类型识别方法

Interference Type Classification Method Based on Spatial- Temporal Aware Dual-Stream Fusion Network

摘要:

针对复杂电磁环境下各种干扰导致雷达目标识别性能下降的问题，提出一种基于时空感知双流融合网络的干扰类型识别方法。该方法通过构建局部特征提取支路和全局特征提取支路的双流架构，分别捕获距离-多普勒图像的空间局部细节特征和时间全局上下文特征，并设计特征融合模块实现多尺度特征的有效融合。进一步提出分类优化-特征判别损失函数，通过联合优化分类损失和特征判别损失增强模型对各种类型干扰和真实目标的区分能力。实验结果表明，所提方法在复杂电磁环境下显著提升了干扰和目标识别的准确率和鲁棒性，为雷达抗干扰技术提供了新的解决方案。

Abstract:

To address the performance degradation of radar target recognition in complex electromagnetic interference (EMI) environments, an interference type classification method based on spatial-temporal aware dual-stream fusion network (STF-Net) was proposed. The method constructed a two-branch architecture comprising a local feature extraction branch and a global feature extraction branch, which respectively capture spatial-local detail features from range-Doppler images and temporal-global contextual features. A dedicated feature fusion module was designed to effectively integrate multi-scale representations. Furthermore, a classification-optimized feature discrimination loss function was proposed to jointly optimize classification loss and feature discrimination loss, thereby improving the model’s capability to distinguish between various types of interference and real targets. Experimental results demonstrate that the proposed method significantly enhances both accuracy and robustness of interference and target recognition in complex EMI scenarios, offering a novel solution for radar anti-interference applications.