现代防御技术 ›› 2025, Vol. 53 ›› Issue (2): 129-140.DOI: 10.3969/j.issn.1009-086x.2025.02.014

• 目标特性与探测跟踪技术 • 上一篇    下一篇

入射视线角引导雷达图像特征融合的气动目标识别方法

李家宽, 冯博, 申伦豪, 叶春茂, 余继周   

  1. 北京无线电测量研究所,北京 100854
  • 收稿日期:2024-03-16 修回日期:2024-07-05 出版日期:2025-04-28 发布日期:2025-04-30
  • 作者简介:李家宽(1999-),男,河北邯郸人。硕士生,研究方向为雷达目标识别。

Aerodynamic Target Recognition Method Based on Incidence Angle-Guided Radar Image Feature Fusion

Jiakuan LI, Bo FENG, Lunhao SHEN, Chunmao YE, Jizhou YU   

  1. Beijing Institute of Radio Measurement,Beijing 100854,China
  • Received:2024-03-16 Revised:2024-07-05 Online:2025-04-28 Published:2025-04-30

摘要:

高分辨逆合成孔径雷达(ISAR)图像在目标识别中发挥着重要作用,但是获取目标高分辨的ISAR图像需要长时间的雷达照射,无法满足雷达资源调度需求。对此提出了一种入射视线角引导雷达图像特征融合的识别方法,选取驻留时间内的两端时间资源用于成像,满足雷达其余功能在时间上的调整与分配。考虑到不同机型图像调制现象的差异以及同一机型在不同入射视线角下的调制差异,设计混合注意力残差模块和角度引导注意力模块使网络有针对性地关注图像的关键区域并将目标特征与目标姿态进行关联。通过特征融合模块进行图像特征的整合以实现融合识别,最终通过3类飞机实测数据证明,该方法能在满足雷达资源调度需求的前提下获得较高的识别精度。

关键词: 逆合成孔径雷达图像, 目标识别, 卷积神经网络, 注意力机制, 入射视线角, 特征融合

Abstract:

High-resolution inverse synthetic aperture radar (ISAR) images play a crucial role in target recognition. However, obtaining high-resolution ISAR images requires prolonged radar exposure, which does not meet the demands of radar resource scheduling. In this paper, we propose a recognition method that guides radar image feature fusion based on incidence angles. A portion of pulse accumulation imaging during the dwell time is selected to meet the temporal adjustments and allocations of the remaining radar functions. Considering the differences in modulation phenomena among images of different aircraft models and the modulation variations of the same model at different incident angles, a hybrid attention residual module and angle-guided attention module are designed to focus the network on critical areas of the image and correlate target features with target attitudes. A feature fusion module is employed to integrate image features for fusion recognition. Through practical data of three types of aircraft, we demonstrate that this method achieves higher recognition accuracy while satisfying radar resource scheduling requirements.

Key words: inverse synthetic aperture radar(ISAR) image, target recognition, convolutional neural network, attention mechanism, incident angle, feature fusion

中图分类号: