Technique Of Human Micro-doppler Feature Extraction And Tracking With Multi-station Low Frequency Radar

Multi-station low frequency radar can work well in all-weather and all-time monitoring applications.It has the ability to detect the moving target shielded by the foliage and can meet the requirement of security and ecology conservation.The low cost,low hardware complexity and large detection area make it suitable for human monitoring.In this thesis,the foliage penetration human micro-Doppler feature extraction and tracking are studied on the real system.The existing micro-Doppler simulation studies are proposed for mono station radar.The micro-Doppler in bistation radar has relationships of both the moving direction of the human and the bistation angle.The bistation radar cross section(RCS)is different from mono station radar,either.The whole process of the proposed bistation radar microDoppler signal simulation and analysis system is described in detail.The moving trajectories of different parts of the human are simulated by the motion data based human model.A bistation ellipsoid scattering model is proposed to simulate the RCS of different body parts.Processes of signal simulation and processing are given.Finally,the microDoppler analysis are based on the time frequency representation method.Simulated and experimental results demonstrate the validity of the proposed micro-Doppler signal simulation and analysis system.In foliage penetration environments,the components of clutter are very complex.Using traditional micro-Doppler feature analysis method cannot obtain the high quality slow time-Doppler spectrogram(STDS).Traditional micro-Doppler feature analysis method does not consider the micro-Doppler feature separation problem,either.Therefore,a moving trajectory based micro-Doppler feature separation method is proposed to enhance the quality of STDS and separate the micro-Doppler feature of multiple human.The method including several steps.The clutter suppression step and the pulse accumulation step are used to increase the signal to clutter ratio.The one dimensional tracking step is introduced to extract the moving trajectory of the human.Select the range cells which contains the moving information of the human and calculate the STDS.The micro-Doppler feature is estimated by the model fit method.If the range profiles of the human are not overlapped,the proposed method can perfectly obtain high quality STDS and separate the micro-Doppler feature of different human.The existing range based minimum mean square error paring(MMSEP)multistation multi-target localization(MSMTL)method only uses the mean square error(MSE)of the range for measurement pairing.It has a large probability of wrong pairing and affects the location accuracy.A range-Doppler(RD)based MSE metric is proposed for measurement pairing.The S-dimensional assignment(SDA)algorithm is also introduced for RD measurement pairing.Simulation results reveal that the location accuracy has the relationship with pairing metric,pairing algorithm,measurement errors,missing detections and false alarms.The RD based MSE metric overcomes the range and the velocity MSE metric in MSMTL.The SDA pairing method overcomes the MMSEP in MSMTL.The RD based MSE metric by SDA pairing MSMTL has the best location performance.The track accuracy of different range measurement estimation methods has not been evaluated in recent studies.For extended human targets,the modified geometrical center(GC)and maximum impulse response amplitude(MIRA)methods are proposed for the range measurement estimation.The experimental results demonstrate the validity of the GC and MIRA range measurement estimation methods.Traditional human tracking methods detect and estimate the measurements on the range profiles.If the human targets overlap on the range profiles,it may miss the measurements and have inaccurate measurement estimation which lead to the discontinuities of the final tracking trajectories.A RD based human tracking method is proposed to improve the tracking performance.The Doppler information provides the other dimensional resolution capability.By detecting and estimating the measurements on the instantaneous RD image,it can decrease the missing measurements and inaccurate measurements which lead to less missing and wrong locations.The experimental results demonstrate the advantage of the proposed human tracking method.