Research On Micro-drones Classification Technology Based On Micro-doppler Radar

Micro-drones not only brings convenience to people's life,but also causes a series of social problems,so it is urgent to classify different types of drones accurately.The radar micro-drones identification is mainly realized by the micro-Doppler effect generated by the rotating rotor.Compared with the traditional identification method,this method has the advantages of wide application and free from weather interference.Based on the micro-Doppler feature recognition process,a new time-frequency analysis method and a feature extraction algorithm are proposed in this paper.The main results of this study are as follows:(1)The micro-Doppler effect in radar is introduced,the special micro-Doppler echo of rotating rotor is analyzed in detail,the mathematical model of the echo signal of rotor blade was established,and the theoretical echo signal and its time-frequency analysis results were simulated in the MATLAB software environment.The practical echo data acquisition of three kinds of micro-drones rotors is carried out by using portable radar equipment.(2)In the process of time-frequency analysis of the echo signal,a new time-frequency analysis method based on synchrosqueezing short-time fourier transform(SSTFT)is proposed.SSTFT can effectively improve the aggregation of time-frequency curves by compressing spectrums with short-time fourier transform(STFT)in the frequency dimension.Through the verification of the actual data,it is found that compared with the classical time-frequency analysis method,SSTFT has high time-frequency resolution and can also reconstruct the signal,which is conducive to the effective recognition of the micro-drones.(3)In the feature extraction of echo signal,a multi-dimensional feature extraction algorithm is proposed.This algorithm mainly extracts time-frequency feature and frequency-change feature from the time-spectrum map obtained by SSTFT.The three-dimensional entropy of the time-frequency feature,the frequency-change feature is characterized by the position of Cadence-Frequency Spectrogram(CFS)peak and the mean square deviation of CFS after removing the peak.The comparison experiment of Support Vector Machine(SVM)classification accuracy shows that the proposed algorithm enriches the characteristic dimension of time-spectrum and improves the recognition accuracy.