<p>The initial and essential step in Electronic Support (ES) systems is the deinterleaving of multiple radars operating simultaneously in an electronic warfare (EW) environment. The deinterleaving of multiple advanced radar signals, having the capability of varying carrier frequencies (RF), pulse widths, and pulse repetition intervals (PRI), is a challenging problem. This paper proposes an instance segmentation based deinterleaving method that utilizes amplitude patterns generated on the ES system, which are directly related to the types of radar antenna scans. The method offers robustness against variations in both RF and PRI types and values. The pulses corresponding to each radar signal are identified using the masks produced by the instance segmentation method. To the best of our knowledge, this is the first study to achieve real-time deinterleaving using amplitude and time of arrival parameters with an instance segmentation model considering all known radar scan types. The effectiveness of the proposed method is evaluated through simulations that consider varying numbers of radar signals and missing pulse rates. Even under the most challenging conditions with five radar signals and a 30% missing pulse rate, the method achieves an F1-score approaching 0.90. Furthermore, real-time deinterleaving is validated using emulated radar pulses.</p>

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

Radar signal deinterleaving using deep learning based instance segmentation

  • Mehmet Burak Kocamış,
  • Adnan Orduyılmaz,
  • Selçuk Taşcıoğlu

摘要

The initial and essential step in Electronic Support (ES) systems is the deinterleaving of multiple radars operating simultaneously in an electronic warfare (EW) environment. The deinterleaving of multiple advanced radar signals, having the capability of varying carrier frequencies (RF), pulse widths, and pulse repetition intervals (PRI), is a challenging problem. This paper proposes an instance segmentation based deinterleaving method that utilizes amplitude patterns generated on the ES system, which are directly related to the types of radar antenna scans. The method offers robustness against variations in both RF and PRI types and values. The pulses corresponding to each radar signal are identified using the masks produced by the instance segmentation method. To the best of our knowledge, this is the first study to achieve real-time deinterleaving using amplitude and time of arrival parameters with an instance segmentation model considering all known radar scan types. The effectiveness of the proposed method is evaluated through simulations that consider varying numbers of radar signals and missing pulse rates. Even under the most challenging conditions with five radar signals and a 30% missing pulse rate, the method achieves an F1-score approaching 0.90. Furthermore, real-time deinterleaving is validated using emulated radar pulses.