A Novel Method for Sea Clutter Suppression and Target Detection via Deep Convolutional Autoencoder

In this paper, we investigate target detection based on the different relevance of sea clutter with a deep learning approach. The proposed method employs deep convolution autoencoder (CAE) to learn the necessary features and classification boundaries using the simulated data without employing any explicit features on the pulsed radar signals. Compared with conventional methods for sea clutter suppression, our algorithm do not need to estimate the covariance matrix of clutters. Specifically, we can automatically remove complex patterns like superimposed clutter from a target, rather than simple patterns like pixels missing at random. The results show that the proposed deep learning approach has very reliable detection performance compared with space-time adaptive processing (STAP), even at low signal-to-clutter ratios. Key–Words: convolutional auto-encoder, deep neural network, sea clutter suppression, target detection.