Recognition of Drone Formation Motions Based on Radar Cross Section Statistical Characteristics

‍ ‍Micro-drones with varying formations have garnered significant attention for applications in low-altitude airspace. This study focuses on the characterization and classification of formation motions in micro-drones. It analyzes the mean value， linear frequency spectral coefficient （LFSC）， Mel frequency cepstral coefficient （MFCC）， and modified MFCC of radar cross section （RCS） sequences corresponding to different micro-drone formations， along with their statistical distribution characteristics. We propose a modified hidden Markov model （HMM） that incorporates prior knowledge of feature distributions to classify these formations， enabling the motion state classification of micro-drone formations using limited training data from a single micro-drone. Experiments were conducted in the X band， utilizing both HH and VV polarizations， to assess the effectiveness of the proposed method. The classification results achieved an average accuracy of over 89%， demonstrating the model’s capability in differentiating between various formations and numbers of micro-drones.