Radar Weak Target Detection Method Based on Deep Manifold Network

Target detection in strong clutter is an important task in radar signal processing. In this paper, a weak target detection method based on deep manifold network is proposed. Specifically, the time-frequency features of the radar echo, extracted by the short-time Fourier transform and Wigner-Ville distribution, are constructed as symmetric positive definite matrices. After multiple dimensionality reduction of the matrices in the deep manifold network-SPDNet, the target and clutter are distinguished in different regions of the manifold, which improves the detection performance of weak radar targets. Finally this paper carries out experiments based on publicly available measurement datasets to effectively validate the detection performance of this manifold network.