Journal of Infrared and Millimeter Waves, Volume. 43, Issue 6, 775(2024)

DIFNet: SAR RFI suppression network based on domain invariant features

Wen-Hao LV2,3, Fu-Ping FANG1、*, and Yuan-Rong TIAN1
Author Affiliations
  • 1School of Electronic Science,National University of Defense Technology,Changsha 410073,China
  • 2University of Chinese Academy of Sciences,Beijing 100049,China
  • 3School of physics and optoelectronic Engineering,Hangzhou Institute for Advanced Study,University of Chinese Academy of Sciences,Hangzhou 310024,China
  • show less
    Figures & Tables(14)
    Flow chart of RFI suppression network based on DIFNet
    DIFNet diagram
    Internal structures of DIFNet
    SAR imaging model
    Iterative loss curve of the network
    Time frequency spectrogram:(a)interfered time-frequency spectrogram in same-sensor experiment;(b)time frequency spectrogram by notch filtering;(c)time frequency spectrogram by PISNet;(d)the time-frequency spectrogram by DIFNet
    Suppression results in MiniSAR:(a)interfered image;(b)label;(c)result by notch filtered;(d)result by PISNet;(e)result by DIFNet
    Time-frequency spectrogram in cross-sensor experiment:(a)interfered time-frequency spectrogram;(b)notch filtering result;(c)DIFNet filtering result
    Cross-sensor results in Sentinel-1:(a)interfered image,ME=3.40;(b)notch filtering image,ME=2.34;(c)DIFNet filtering image,ME=2.19
    • Table 1. DIFNet’s pipeline

      View table
      View in Article

      Table 1. DIFNet’s pipeline

      Algorithm I:DIFNet’s pipeline

      1. Detect RFI in SAR images;

      2. Perform STFT pulse-by-pulse;

      3. Predict RFI by DIFNet;

      4. Subtract RFI;

      5. Perform ISTFT pulse-by-pulse;

      6. Convert SAR echoes into SAR images.

    • Table 2. Radar parameters of training and testing Data

      View table
      View in Article

      Table 2. Radar parameters of training and testing Data

      Data

      Parameters

      Training data

      Same-sensor

      experiments

      Cross-sensor

      experiments

      SourceMiniSARMiniSARSentinel-1A
      BandX-bandX-bandC-band
      Bandwidth1.5 GHz1.5 GHz100 MHz
      Polarization modeHHHHVV/VH
    • Table 3. Interference parameters in MiniSAR

      View table
      View in Article

      Table 3. Interference parameters in MiniSAR

      Bandwidth

      Parameters

      NarrowbandBroadband
      Interference bandwidth<30 MHz30 MHz~150 MHz
      SIR-15 dB~0 dB-15 dB~0 dB
      Interference source23~5
    • Table 4. Same-sensor Result

      View table
      View in Article

      Table 4. Same-sensor Result

      Methods

      Indicators

      Interfered imageNotch filteringPISNetDIFNet
      PA/94.95%/96.55%
      IoU/74.41%/80.40%
      PSNR/dB11.3621.4925.0523.51
      ME0.240.090.040.04
    • Table 4. Cross-sensor Result

      View table
      View in Article

      Table 4. Cross-sensor Result

      Parameters

      Indicators

      Interfered imageNotch filteringDIFNet
      PA/92.67%94.56%
      IoU/78.63%81.23%
      ME3.402.342.19
    Tools

    Get Citation

    Copy Citation Text

    Wen-Hao LV, Fu-Ping FANG, Yuan-Rong TIAN. DIFNet: SAR RFI suppression network based on domain invariant features[J]. Journal of Infrared and Millimeter Waves, 2024, 43(6): 775

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Millimeter Wave and Terahertz Technology

    Received: Mar. 26, 2024

    Accepted: --

    Published Online: Dec. 13, 2024

    The Author Email: Fu-Ping FANG (capkoven@mail.ustc.edu.cn)

    DOI:10.11972/j.issn.1001-9014.2024.06.008

    Topics