Space Target Recognition Based on RCS Feature Extraction Using Relative Position Matrix

Space target recognition plays a crucial role in ballistic missile interception, especially in the vital task of distinguishing between warheads and decoys. This paper proposes a method to transform one-dimensional (1D) radar cross section (RCS) series into two-dimensional (2D) images by utilizing global information through Relative Position Matrix (RPM). Additionally, the ConvNeXt is enhanced by incorporating a random noise module and introducing a noise reduction self-coding layer to bolster the model's resistance to signal interference. The experimental results indicate that the proposed method achieves high accuracy in recognizing missiles.