Radar Signal Sorting Based On Improved Limited Penetrable Visibility Graph Network

In modern electronic warfare,radar signal sorting is an indispensable part.In the field of radar signal sorting,it is necessary to correctly sort the received radar mixed pulse signals to ensure that the subsequent processing of reconnaissance receivers can proceed smoothly and obtain radar pulse signals belonging to the same radar radiation source.With the complex and diverse electromagnetic pulses in the electromagnetic environment and the continuous development of radar systems,traditional sorting techniques face enormous challenges.For example,when the modulation parameters of multiple radars are similar,radar signal sorting may regard them as the same radar,resulting in a "missed batch" phenomenon.In addition,multi-functional radars in modern radar systems can be sorted into multiple radars for the same radar,resulting in an "increased batch" phenomenon due to switching different operating modes for different tasks.To improve the performance of multi-functional radar signal sorting in complex electromagnetic environments,this thesis combines complex networks with radar signal sorting technology to conduct research,with the following main work:1、Analyze the current research status of traditional radar signal sorting and complex networks,introduce the commonly used characteristic parameters and conventional sorting algorithms of radar signal sorting,and provide a detailed introduction to the basic knowledge of complex networks.2、Aiming at the poor sorting effect of multi-functional radar signals,an improved limited penetrable visibility graph(ILPVG)algorithm is proposed,which firstly establishes a network model using visualization criteria and DOA difference criteria,and analyzes the network model based on the topological properties and centrality index of complex networks.Then,the community structure is detected through a label propagation algorithm and density peak clustering algorithm to obtain the radar signal sorting results.Simulation experiments show that in the multi-functional radar sorting scenario,the proposed method improves the sorting performance by 3.46% compared to using the LPVG complex network sorting method.Moreover,the sorting accuracy of this algorithm is significantly better than the k-means-based sorting method and the DBSCAN-based sorting method,with sorting performance improved by 11.875% and 11.25%,respectively.3、In view of the problem that the traditional graph neural network model does not consider the structural information of nodes in the network when classifying network nodes,A radar signal sorting method based on an improved Graph SAGE network is proposed.First,the radar pulse signal sequence is transformed into a complex network model using the ILPVG algorithm to obtain preprocessed data.Then,the corresponding feature matrix and adjacency matrix of the complex network is used as the input of the Graph SAGE network.During mini-batch training of the improved Graph SAGE model,nodes are obtained based on the clustering coefficient in descending order to form the training node-set,and part of the nodes are trained each iteration until the model converges.Simulation experiments show that compared with the sorting algorithms based on the GCN network and the Graph SAGE network,the proposed method improves the average sorting accuracy by 14.95%and 8.3%,respectively.Compared with the sorting method in Chapter 3,the average sorting accuracy of radar signal sorting is improved by 4.17%.