Study On Feature Extraction And Classifier Design Of Airplane Targets Based On Narrowband Radar

When radar illuminates a moving target,the carrier frequency of the returned signal will be changed,which is called as Doppler effect.In general,there are also some moving components of the relative main movement on the target,such as aircraft rotor,turbofan engine blades,wheels.These parts may induce the additional Doppler modulations around the target's Doppler frequency,which is known as micro-Doppler effect.Micro-Doppler modulation contains the configuration structure and the motion characteristic of the target,which can be used for aircraft classification.The main contents of this dissertation can be summarized as the following three aspects:1.Firstly,we introduce an ideal parametric model of the return of aircraft.Based on the simulated data generated by the model,we analyze the radar parameters of the return.To cope with the radar antenna pattern,an adaptive demodulation method is proposed.The results based on the measured data show that the method can eliminate the modulation effectively.Finally,the simulated signal echoes of the three types of aircraft are analyzed,and it is found that there are obvious differences in the micro-Dopple components of the echoes of the three types of aircraft.2.A feature extraction method based on Complex Local Mean Decomposition(CLMD)is proposed.This method decomposes the echo signal to a series of component signal based on CLMD at first.After that,six kinds of features of micro-Dopple components are extracted to be used.Experiment results show that the extracted can be used to classify the three types of aircraft effectively.3.To cope with some problems of support vector machine classifier(SVM),such as parameter optimization,a classification method based on random forest(RF)is proposed.The experimental results show that RF is superior in generalization performance,computational complexity and parameter sensitivity.Compared with the other method,the rationality of the random forest classifier to the feature evaluation is analyzed and verified.On this basis,a feature selection method is proposed.The experimental results based on the simulated and the measured data show that the method can select the better feature set with moderate dimension and guarantee the classification performance.