Classification of Bird and Drone Based on Radar Time-Frequency Analysis

To address the classification of 'Low, Small, Slow' targets such as birds and drones, this paper proposes a classification algorithm based on target time-frequency spectrogram analysis. Firstly, radar-collected data is subjected to Short-Time Fourier Transform (STFT) to obtain target time-frequency spectrograms, followed by data preprocessing. Subsequently, these time-frequency spectrograms are fed into the designed neural network Residual Pyramid pooling Net (RPNet) for the classification of the two target classes. Validation results based on radar experimental data demonstrate that this algorithm can effectively classify birds and drones, thereby improving the classification performance of small targets.