Siamese Sparse Reconstruction Network for Ionospheric Clutter Suppression in HFSWR

High-frequency surface wave radar (HFSWR) suffers from performance degradation due to strong ionospheric clutter. In this article, we propose a deep learning-based method, siamese sparse reconstruction network (SSRN), for ionospheric clutter suppression in HFSWR. The SSRN integrates sparse coding into a siamese encoder-decoder structure tailored for complex-valued radar data. The convolutional layers perform iterative approximation of the range-Doppler spectrum degradation function. The RD spectrum reconstruction loss preserves phase characteristics. The enhancement in signal-to-clutter-and-noise ratio gain within the ionospheric clutter using the wavelet series oblique projection filter suppression algorithm is observed to be approximately 4 dB. Experiments demonstrate the advantages of SSRN in clutter suppression over traditional sparse reconstruction and deep learning methods. This data-driven approach combining deep learning and sparse coding holds promise for enhancing HFSWR clutter mitigation.