Answers are grounded in this paper and cite their sources.
The rapid and accurate distinction between ship targets and corner reflectors is a research hotspot in radar target classification, and existing methods based on microwave scattering characteristics have achieved good performance. However, when faced with low Signal-to-Noise Ratio (SNR) and scenarios with limited labeled samples, the accuracy of existing methods is significantly compromised. This performance degradation essentially stems from insufficient mining and utilization capabilities of microwave scattering characteristics under challenging conditions. To solve this problem, this paper proposes a target classification method based on agile frequency scattering coefficient sequences and complex-valued autoencoder. Firstly, the proposed method performs polarization rotation on each target’s multi-frequency scattering coefficient samples, thereby generating the Frequency-Polarization Rotation Scattering Coefficient (FPRSC) matrix. Secondly, a Complex-valued Autoencoder with Attention (CVAE-A) is designed to achieve noise reduction, in which complex-valued processing greatly enhances feature extraction performance. Finally, the pre-trained encoder from CVAE-A is employed as a feature extractor. Combined with several fully connected layers, it forms a semi-supervised classifier that can be effectively trained using limited labeled samples. Experiments on both simulated and measured data demonstrate the method’s dual capability: achieving robust denoising of target scattering coefficients in low-SNR environments while maintaining accurate classification performance with limited labeled training samples.
No takes yet. Share an insight, caveat, or question.
Lv et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: