RCS Statistical Feature Extraction for Space Target Recognition Based on Bi-LSTM

Space target recognition is of great significance for ballistic missile interception, especially in distinguishing between real and decoy warheads. This paper proposes a recognition method that combines feature extraction and Bi-LSTM, surpassing existing algorithms in classification performance. To obtain more discriminative sequences, the method extracts 11-dimensional statistical features of the Radar Cross Section (RCS) sequence using a sliding window, and inputs the feature sequence into the Bi-LSTM network for classification. Even under low signal-to-noise ratio conditions, the proposed algorithm maintains strong performance, demonstrating its noise resistance and generalization ability.