HFSWR Clutter Recognition Based on Attention DCNN

High frequency surface wave radar (HFSWR) suffers from passive interference caused by ionospheric clutter. Ionospheric clutter has a chaotic nature that makes a priori information insufficient for clutter suppression algorithms. In recent years, various clutter recognition methods for suppression have been proposed, including machine learning methods such as SVM, CNNs, and LSTM. However, existing models still have poor generalization performance and struggle to sustain high multi-class recognition accuracy in a realistic, time-varying, and highly complex ionospheric clutter environment. To address this issue, this paper proposes a deep learning ionospheric clutter recognition model with expansive convolutional networks. The model replaces the traditional convolutional-full-connected network with a modified dilated convolutional network with attention mechanism (AM-DCNN) to improve feature resolution and sample size requirements while maintaining recognition accuracy. The proposed method achieves high accuracy in recognizing five refinement characteristics of strong ionospheric clutter.