Research On Classification Of Airplane Targets Based On JEM Effect

In recent years, micro-Doppler effect is widely used in radar target classification and recognition. The micro-Doppler characteristics can reflect the motion characteristics and geometry of the target, therefore, the differences of the micro-Doppler characteristics can be used to classify targets. The periodic modulation induced by rotors of helicopter, propeller and jet aircraft on the radar echo, is called Jet Engine Modulation(JEM). Three categories of aircraft are classified by their different micro-Doppler characteristics.The main contents of the paper can be summarized as the following three parts:The first part, based on the previous work, we establish the radar echo parameter model of horizontal rotor and the radar echo parameter model of vertical rotor, respectively; By utilizing simulation data based on the radar echo parameter model and simulation data obtained by electromagnetic calculation of FEKO software, we compare and analyze the differences of JEM effects of helicopter, propeller and jet aircraft rotor.The second part, we introduce several existing feature extraction methods in time domain, Doppler-frequency domain and eigenvalue spectral domain, then the paper propose a new feature extraction method based on time-domain echo, called the time-domain correlation features, and the extracted new features can reflect periodic characteristics of time-domain echo. Based on simulation data and measured data, it is proved that the time-domain correlation features require a lower pulse repetition frequency, that is, the classification performance of features extracted in Doppler-frequency domain degrade when there is a aliasing in Doppler-frequency domain, but the classification performance of time-domain correlation features can still perform relatively well.The third part, through azimuth domain division and feature selection, we can enhance the classification accuracy of three categories of aircraft. In this section, firstly, with the altering of azimuth, we analyze the change of the echo modulated by three categories of aircraft rotors and the change of their corresponding characteristics, and we propose amethod of dividing azimuth domains, in order to reduce the distribution of characteristics in the same domain; secondly, we introduce the basic principle of linear RVM, and the feature selection method based on linear RVM; Finally, experiments confirm that the azimuth domain division can enhance the classification accuracy of three categories of aircraft, then through selecting features in different domains, classification accuracy can be further improved.