TransRCSnet: Self-Attention Network for Classifying Space Target Using RCS Time Series
Lijin Fang, Qitong Chu, Ruofan Wang, Yi Min Yang, Yufeng Yue · 2024
The accurate and rapid identification of non-cooperative warheads and decoys is of great significance to ensure national security. Traditional identification methods, while achieving high accuracy through feature extraction and machine learning, are often constrained by reliance on human expertise and struggle with nonlinear data in complex scenarios. To address these limitations, we propose TransRCSnet, an innovative recognition network that leverages the capability of time series neural networks to automatically extract high-dimensional features directly from radar cross-section (RCS) time series data. Experiments between TransRCSnet and three conventional classification algorithms are conducted, and the results show that TransRCSnet achieves the best performance for classification and prediction speed in complex scenarios.