Research On Radar Signal Sorting Technology In Complex Environment

In recent years,with the continuous appearance of new radar systems,the electromagnetic environment becomes more and more complex,showing the characteristics of dense and complex signals,variable parameters and serious spectrum overlap.As an important link in electronic reconnaissance processing,radar signal sorting is directly related to the subsequent identification effect and interference decision.Therefore,the research of radar signal sorting technology in complex environment has important theoretical significance and engineering application prospect.Starting from the processing framework of electronic reconnaissance system,this paper introduces the radar signal sorting module,and summarizes the overall working process of radar signal sorting,including feature parameter extraction,clustering pre-sorting,PRI primary sorting and batch processing,radar emitter characteristics analysis and so on.The research is mainly conducted from the following three aspects:Firstly,the correlation algorithm of cluster pre-sorting based on pulse descriptive word is studied to dilute the pulse stream.In this paper,an improved K-Means clustering algorithm is firstly proposed,which does not need to initialize the number of clusters and has high efficiency,but has certain requirements on the dynamic range of clustering parameters.Then the principle of DBSCAN clustering algorithm is briefly introduced,and the DBSCAN clustering algorithm is improved.The algorithm can adaptively calculate neighborhood radius and dynamically adjust the minimum density point threshold.From the clustering results,The adaptive DBSCAN clustering algorithm is free from manual intervention and can still maintain good clustering results in scenarios with wide parameter variation range.Secondly,relevant processes and algorithms based on PRI primary sorting are studied to complete PRI estimation and pulse retrieval of pulse sequences.In this paper,the traditional known radar primary sorting algorithm and typical unknown radar PRI primary sorting algorithm are analyzed,and the application scenarios,advantages and disadvantages of each algorithm are analyzed through experiments.Secondly,a combined batch processing algorithm is proposed to solve the problem of batch addition or sorting failure caused by the fast frequency conversion radar signal.Finally,the convolutional autoencoder network is applied to the primary sorting of known radar signals,and the effects of TOA measurement error and pulse loss rate on the sorting results are investigated experimentally.Experimental results show that the radar signal sorting algorithm based on convolutional autoencoder has higher operational efficiency than the traditional known radar primary sorting algorithm and better sorting effect in the case of interference signal and serious pulse loss.Finally,the characteristics of the sorted radar sequence are analyzed to update the radar parameters and change rules in the database.This paper mainly focuses on signal frequency type identification and PRI modulation type identification.Since both of the two types show periodic change rules in time sequence and the change rules are basically similar,this paper constructed MLP network and CNN network respectively to identify the frequency type and PRI type of radar signals.Experimental results show that the recognition method based on CNN network has better recognition effect.