The Detection And Identification For The Low-altitude And Slow-speed Small Targets

Due to the signatures of low-altitude and slow-speed small targets,the reliable detection and identification is difficult to be achieved by conventional radar systems and detection techniques.As a typical low-altitude and slow-speed small target,small unmanned aerial vehicles(UAVs)are widely applied in military and civilian fields.Micro-motion is defined as a mechanical vibrations or rotations or other high-order motions of an object or any structure comment of the object in addition to its bulk translation.It provides a specific characteristic of the structure to the target,and is paid much attention in the radar-based target recognition area.The rotation property,as the vital signature of a rotor blade,would increase the reliability for the detection and identification of small UAVs.However,the difficulty for the detection and identification of low-altitude and slow-speed small targets is to separate the weak micro motion signal from the returned signal of bulk translation and strong noise.Parameter estimation of micro motion with the time-frequency(TF)analysis is facing some technical bottlenecks.Based on the detection and identification of small UAVs,this dissertation focuses on the parameter estimation of micro motion and detailly investigates the signal model,the separation and reconstruction of the micro motion signal,the location and identification in complex scene,and the feature analysis of micro motion in multichannel radar system.Firstly,the signal models are established based on the ideal scatter center model and geometrical theory of diffraction(GTD)model,respectively.Then,the modulation of micro-Doppler(m-D)frequency is analyzed.In GTD model,the dependance of the scattering properties on the azimuth angle and the frequency of the transmitted wave is investigated.Moreover,since the phase noise,the vibration during the flying,and white Gaussian noise are considered in the m-D signal,the signal model is closer to the physical scene.Secondly,the reconstruction of m-D signal based on the empirical mode decomposition(EMD)is achieved.EMD is a data dependent method of decomposing a signal into a series of intrinsic mode functions(IMFs)by their characteristic frequency scales in the data empirically.The distribution of extrema could reflect the local behavior of the data on the frequency,which is similar to that of the relevant components in the input signal.In this case,the periodical sine wave is constructed as the approximated Doppler signal with the obtained parameters from the corresponding IMF.Then,this dissertation develops the reconstruction methods of m-D signal based on the masking signal model and sparse recovery,respectively.The micro motion frequency,which is estimated from the reconstructed m-D signal,would be the significant signature for the identification of small UAVs.Taking into account the actual measurement conditions,the detection and identification methods based on the theory of the correlation are investigated.A series of spectrograms of the returned signal are regarded as optical images to establish the Gaussian mixture model(GMM).Conceptually,since the frequencies induced by moving vehicles are exclusively shifted regularly with the change of the distance to the radar over the different periods,the spectrum induced by moving vehicles could be extracted as the foreground and the rest would be classified as the background,Following the periodical iteration,the motion state of each vehicle will be updated,and the real-time tracking parameters are acquired from the characteristic curve on the spectrograms via Hough transform(HT).On the other hand,due to the object with the similar micro motion frequency in the detection scene,this dissertation proposes the cyclostationary phase analysis(CPA)to acquire the micro motion frequency and amplitude.In this case,the false alarms with the similar micro motion frequency could be eliminated,and the reliability for the identification is improved.Finally,this dissertation investigates the multivariate synchrosqueezing transform(SST)based on the multichannel radar to improve the concentration of TF ridges.Since the basis function of SST is highly matched with the form of the m-D modulation,the proposed method enhances the concentration of TF representation for m-D signal,but also reduces the TF energy for Doppler signal and multi-frequency false alarms.In order to improve the immunity of noise,the noise assisted multivariate empirical mode decomposition(NA-MEMD)is combined with the multivariate SST.Then,the purely m-D signal is decomposed without the contamination of the interference frequencies.To summarize,this dissertation focuses on the signal model,location and identification,the application on the multichannel radar to intensively investigate the detection and identification techniques of small UAVs,proposes a series of m-D signal processing methods,which have been verified in field experiments.These research results bear certain significance for the acceleration of the application to detect and identify small UAVs.