Sea Clutter Amplitude Prediction via an Attention-Enhanced Seq2Seq Network

Sea clutter is a kind of ubiquitous interference in sea-detecting radars, which will definitely influence target detection. An accurate sea prediction method supposed to be beneficial while existing methods are based on the one-step-ahead prediction. In this paper, network (SCPNet) proposed achieve k-step-ahead characteristics clutter. The SCPNet takes sequence-to-sequence (Seq2Seq) structure as backbone, and simple self-attention module employed enhance ability adaptive feature selections. normalized amplitudes inputs capable predicting an output sequence with length k; phase space reconstruction theory also used find optimized input sequence. Results radar data-sharing program (SDRDSP) database show mean square error 1.48 × 10−5 8.76 10−3 eight-step-ahead prediction, respectively. Compared four methods, achieves best performance.