Radar Aircraft Target Recognition Based On Jet Engine Modulation

In modern warfare,there are many kinds of aircraft targets,and different types of aircraft have different functions and threats.It is of great significance for radar to recognize targets after detecting targets.In the course of flight,the rotating parts of the aircraft will periodically modulate the radar signal.This modulation characteristic is called jet engine modulation(JEM).The narrowband radar target recognition can be realized by extracting the corresponding JEM features from the difference of radar echo modulation characteristics of the aircraft target.The main contents of the thesis can be divided into three parts:In the first part,the principle of airplane radar echo modulation is introduced,and the mathematical model of airplane target radar echo is established based on the principle.Then,the simulation analysis of airplane radar echo signal is carried out according to the mathematical model,and the influence of radar parameters on radar echo modulation characteristics is discussed.In the second part,the classification of aircraft targets based on JEM features is studied.Firstly,several existing JEM feature extraction methods are introduced.Then,according to the difference of normalized amplitude distribution of aircraft radar echo signal,the kurtosis skewness feature of aircraft radar echo is creatively extracted from the point of statistical analysis.Combined with wavelet analysis and linear discriminant analysis,the radar aircraft target classification is realized.The simulation results show that this method has strong anti-noise ability,and does not require high pulse repetition frequency and observation time.It is a good classification feature for aircraft targets.The third part is the research of aircraft target type recognition.Firstly,the popular manifold learning in the field of target recognition in recent years is briefly introduced.Then,according to the characteristics of airplane radar echo,the neighborhood preserving projection(NPP)algorithm which is more suitable for JEM feature extraction is introduced,and the problem of "out of sample" is analyzed.On the basis of this algorithm,the concept of "dissimilarity" is introduced,and the algorithm is further extended to the non-linear form by "kernel",and the enhanced kernel neighborhood preserving projection(EKNPP)algorithm is obtained.At last,the effectiveness and recognition performance of EKNPP algorithm are analyzed through simulation experiments.The results show that EKNPP is a very effective JEM feature extraction method and can get good recognition results.