Research On Clustering And Sorting Algorithm Of Non-collaborative Radar Emitter Signals

Radar countermeasure reconnaissance is the main method to obtain radar countermeasure intelligence.One of its core technologies is to accurately classify the radar radiation source.Only on this basis,can the radiation source identification and threat level judgment be conducted,so that the commanding personnel can choose appropriate confrontation strategy.With the rapid development of electronic warfare,it becomes more and more difficult to predict radar electromagnetic environment.Non-cooperative radar radiation sources with complex and variable feature parameters have gradually become dominant.This thesis mainly focuses on the critical theoretical problems of non-cooperative radar signal sorting with spatial aliasing of feature parameters.The paper analyzes and summarizes the research status of domestic and foreign radar signal sorting fields.The main research direction of this paper is the clustering analysis method in data mining,and two classical clustering algorithm K-Means and fuzzy C-means(FCM)based on division are studied.Radar signals are susceptible to noise pollution and measurement errors,which directly or indirectly leads to an increase in the probability of misinterpretation of sorting processing.Therefore,this paper introduces the concept of data field into radar signal data preprocessing,and proposes a radar signal preprocessing method based on data field.The data field potential value is used to judge the noise and outlier interference in radar emitter signal samples so as to eliminate data outliers.In consideration of the shortcomings of traditional radar signal sorting model and the problems of clustering algorithm for non-cooperative radar signal sorting,this paper constructs a new sorting model structure for non-cooperative radar signals.The initial value of the radar radiation source is determined through the construction of the equipotential line of the data field,so the initial parameter setting step that requires prior information of the radiation source signal can be avoided.Fuzzy C-means clustering sorting algorithm based on soft partition can better deal with class boundary problem of spatial overlap of feature parameters.Consider the combination of data field and fuzzy C-means clustering based on softening.This also solves the problem of class boundary division and cluster initial value setting for feature parameter overlap.