Radar Emitter Recognition Based on SIFT Position and Scale Features

Abstract

As the complexity of the battlefield electromagnetic environment has increased, it has become challenging to identify radiation sources using traditional radar emitter identification technology in low signal-to-noise ratio (SNR) environments. We propose a method of radar emitter recognition based on the scale-invariant feature transform (SIFT) position and scale features. The proposed method first uses time-frequency analysis to analyze the radar emitter signal, and the obtained 3-D time-frequency image is normalized into a 2-D gray image. Next, the Gaussian difference pyramid of the gray image is established, and the SIFT scale features and position features of the image are extracted based on the Gaussian difference pyramid. The extracted noise feature points are suppressed based on the scale features. Finally, a support vector machine is used for the automatic identification of radiation sources based on the SIFT position features. The simulation results show that the proposed method performs well in a low-SNR environment.