A Rotor Target Recognition Method Based on Transfer Learning

Aiming to overcome the limitations of traditional manual search for identifiable annotation features that require professional domain knowledge and difficult to extract high-quality features, the VGG16 model-based transfer learning method is used to automatically extract and identify rotor-type targets in this paper. Firstly, based on the radar echo model of the rotor targets, the echo signals of five typical single-rotor or dual-rotor targets are obtained, and then five typical rotor targets data sets, that is AH-64, K-50, K-MAX, V-22 and WZ-9, are established. Finally, the precise identification of the rotor targets is achieved by fine-tuning the VGG16 model. The simulation results validate that the proposed method can improve the recognition rate of rotor target to nearly 99.76%.