A Novel Sea Clutter Suppression Method Based on Deep Learning with Exploiting Time-Frequency Features

Due to the nonstationary environment, observation system parameters and the low observable characteristics of sea surface targets, the existing sea clutter suppression methods cannot obtain satisfactory results to improve the sea surface target detection performance. In this paper, by exploiting the multi-dimensional domain complementary features of the received echo signal, a novel sea clutter suppression method based on deep learning (DL) with utilizing time-frequency characteristics is proposed. Firstly, the discrete wavelet transform (DWT) is adopted to decompose the echo signal into different sub-bands, in which target and clutter are contributed to the corresponding sub-band. After that, the LeNet-5 neural network is developed to classify and identify each sub-band signals with fractal features. Finally, the sea clutter suppression is realized by zeroing the corresponding clutter sub-band coefficients. Compared with the existing sea clutter suppression methods, the proposed method has the advantages of recognition accuracy due to using the time-frequency information. The effectiveness and advantage of the proposed method are verified by the measured data of CSIR.