UAV Target Detection Using Wideband Radar

Nowadays,unmanned aerial vehicles(UAVs)are more and more widely used in production and life,and the threat they bring also attracts more and more attention.UAV is a low small slow target with hovering capability.The detection of UAVs shuttling between buildings in the urban environment is difficult: Fixed clutters such as buildings are easy to be confused with hovering UAVs.If these clutters enter into the following tracking and identifying process,it will bring unnecessary burden to the system.Therefore,this paper focuses on the problem of UAV detection in urban environment.We use the wide-band radar with carrier frequency of 35 GHz and bandwidth of 1.2GHz to collect data and study the method of removing clutters in the detecting stage using the micro-Doppler effect to improve the performance of UAV detection in cluttered environment.The main reasearch contents are as follows:First,the micro-Doppler effect caused by UAV propellers is analyzed theoretically,and the correspondence between the propeller rotation and the sinusoidal curves in spectrogram is explained.Through simulation,the relationship between the number of rotor blades and the accumulation time of time-frequency analysis and the spectrogram is analyzed,the reason for the periodic peaks of spectrogram in the frequency dimension of is studied,and the micro-Doppler chatacteristics of UAV eacho are analyzed through the measured data.Second,we use cepstrum to extract micro-Doppler features of UAVs,confirming that cepstrum can effectively extract rotating speed characteristic with the radar system adopted in this paper.An UAV detection method based on cepstrum peak detection with adaptive threshold is proposed.Fixed clutters are removed after the cell-averaging constant false alarm rate detection,and then UAVs can be effectively detected in urban building cluttered environment.Last,we also study the UAV detection method based on machine learning and propose an UAV detection method based on recurrent neural network.Compared with the method using common PCA and SVD feature extraction combined with SVM classification,the UAV detection method based on recurrent neural network has certain advantages in distinguishing hovering UAVs from fixed clutters when the SNR is relatively low.