Research On Radar Target Feature Identification Based On RCS Sequence

Radar target feature identification technology has developed greatly in need of modern information warfare.The research on radar target identification teclhnology and electromagnetic scattering characteristic analysis has aroused wide concern in many countries.Radar Cross Section(RCS)is a metrie that reflects the ability of the target to scatter radar signals.It is a very important feature of radar for target recognition.In this thesis,the radar target feature identification method based on RCS sequence is studied.The research focuses on the complex target RCS estimation based on graphic electromagnetic computing(GRECO)and the target feature extraction method and classifier algorithm based oin RCS sequence.The main work of this thesis has the following points:(l)Tlhe evaluation methods of RCS are analyzed in detail,and the principles,advantages and disadvantages of various RCS estimation methods are summarized.For the estimation of complex target RCS,the method of estimating target RCS based on GRECO algorithm is proposed.The target RCS sequence database is generated by GRECO software to provide data samples for subsequent identification.(2)In order to further improve the radar target feature recognition performance,the target feature extraction method based on RCS sequence is studied.Two target featlre extraction methods are proposed:RCS statistical feature extraction method and feature extraction method based on wavelet transform.The feature extraction method based on wavelet transform is studied.The discrete wavelet transform is performed on the target RCS sequence.The local characteristics of the signal are clearly displayed on the time scale plane,and eleven effective statistical features are extracted,which effectively reduces the feature dimension.Then the classifier algorithm based on RCS sequence is studied.K-nearest neighbor,support vector machine,random forest classifier algorithm and fusion classifier algorithm are introduced.A multi-classifier fusion algorithm based on AdaBoost algoritlhm is proposed,which can effectively improve classification identification.(3)In order to verify that the features extracted by the feature extraction method based on wavelet transform have good separability and the classification effect of the fusion classifier is good,the simulation experiment of classification and identification of adjacent spatial targets is carried out.Firstly,the three types of targets in the adjacentspace:hypersonie vehicle X37B,X43A and Patriot PAC-2 missile are geometrically modeled and the corresponding target RCS database is generated by GRECO software.Then the two extraction methods and the four classifier algoritlhms respectively classify and identify the three types of targets in the adjacent space,and observe and analyze the influence of system noise and meteorological clutter on the target identification.lt is proved by experiments that the features extracted by the feature extraction method based on wavelet transform have better separability and are useful for classifier identification of target classification.The fusion classifier based on AdaBoost algorithm has no noise and no clutter.The classification and identification performance is better than the other three classifiers,and as the noise or clutter strength increases,the radar target recognition is more and more seriously affected.Finally,a radar target feature recognition software was developed to realize the entire radar target recognition process.