Research On Target Classification Based On Micro-doppler Effect Laser Detection

Vibration or rotation of an object or structures on the object exist except for the macro movement of the target.Such micro-motion dynamics may induce frequency modulations on the returned radar signal which generate side-band about target's Doppler frequency shift.The micro-Doppler feature reflecting the intrinsic properties of the target can be extracted.Compared with microwave detection,Lidar detection has strong concealment and anti-noise ability.Short wavelength laser is easier to detect an obvious micro-Doppler shift,thereby enhancing the detection sensitivity and resolution.Laser detection of micro-Doppler effect opens up a new way of feature extraction and target recognition,which has important research and application value.Focusing on micro target classification,a laser coherent detection micro motion echo model is established,and the target vibration and rotation micro Doppler characteristic parameters are extracted,and the classification of typical targets is completed.Specific research includes:Firstly,the principle of laser coherent detection of micro Doppler effect is discussed and the target vibration echo model is established.The rotating echo model of extended target is constructed by physical optics method and the sub-frequency band characteristics different from the traditional linear scattering model are defined and extracted by simulation.According to the characteristics of the micro Doppler echo signal,the suitable time-frequency analysis method is given.Secondly,the target vibration parameters are estimated and verified by experiments.The cyclic autocorrelation function inverse Radon transform(IRT)combined with density centroid algorithm is used to estimate the vibration frequency,amplitude and phase.A frequency correction algorithm based on reconstruction point of IRT extremum searching is proposed.Simulation and experiment results show that the uncertainty of frequency estimation is less than 0.5% while the amplitude estimation error is better than that of micro scale.Besides,the multi component parameters can be estimated effectively with this algorithm.Thirdly,a method of rotating object classification based on texture feature of time-frequency map is proposed.Considering the difference of periodic modulation resulted by typical rotating components,the optimized GLCM and Tamura features are extracted.The simulation shows that these features have strong discrimination for different aircraft targets.Based on the extended target rotation echo model,a micro-Doppler detection and calculation method is presented for the typical rotational target's aspect ratio,root-shoot ratio and the sweep parameter of the propeller tip.Finally,the classification effect of typical targets is tested.The typical target vibration parameters caused by the engine are estimated,then the k-nearest neighbor classifier is used to classify the tank and the vehicle,as well as the aircraft and the missile.The simulation proves that different vibration targets can be distinguished effectively.The features of typical rotor echoes of helicopter,propeller aircraft and jet are extracted by simulation and the result of support vector machine classification shows that the accuracy of target classification can reach more than 98%by using Tamura feature when SNR is over 10dB.Meanwhile,the correct rate is up to 96.4%by using GLCM feature as SNR is 0dB then accurate classification of aircraft targets can be achieved under low SNR conditions.The innovation of this paper includes: The model of the extended target rotation echo is established then the detection and calculation method of the typical rotor shape parameter is given.The correction algorithm of vibration frequency estimation based on reconstruction point of IRT extremum searching is proposed and the multi vibration target detection and parameter estimation are realized by experiment.A classification method of aircraft target based on periodic modulation texture feature is proposed and the result of high classification accuracy under low SNR is obtained.The research results can provide technical support for target reconnaissance and identification of the new system.