HED-CNN based Ionospheric Clutter Extraction for HF Range-Doppler Spectrum

High Frequency Surface Wave Radar (HFSWR) suffers seriously with the ionospheric clutter formed from ionosphere echoes. The ionospheric clutter could be extensive and exists all day long, which restricts the detection performance of HFSWR. It is necessary to eliminate the interference of ionospheric clutter which overwhelms target echoes always. However, there is not a prior knowledge about clutter each work cycle, and anti-ionospheric interference technology adapting to all kinds of situations. With the purpose of extracting the ionospheric clutter separately for clutter cancellation, image processing method is adopted to study and analyze the application of deep learning in edge extraction of ionospheric clutter existing in Range-Doppler (RD) spectrum. In this paper, holistically-nested edge detection (HED) based algorithm is adopted and Canny algorithm is used for comparison. It shows that HED algorithm is effective and efficient in edge extraction of ionospheric clutter in RD spectrum.