Research On Modeling And Analysis Of Electronic Target Sequences

In modern electronic support and confrontation,there are already a large number of mature algorithms for the behavior analysis of a single target in a complex electromagnetic environment,and good results have been obtained.However,the characterization and analysis of the target system as a whole in the real-time environmental scenario is also essential,which is conducive to users to understand the situation of the entire environmental scenario and make timely predictions and decisions based on the analysis results.This kind of scene target that contains information such as targets,relationships,parameters,etc.is called the electronic target sequence as a whole in this thesis.Based on the characteristics of the electronic target sequence,,we focuses on how to carry out the modeling and representation of electronic target sequences and how to analyze electronic target sequences to predict the overall or unknown information of the scene.For electronic target sequences or similar scene data sets,relatively most of the current modeling methods are for command system modeling,or simply for individual modeling such as ships and aircraft.Cross-level information,overall comprehensive information and overall dynamic information,can not be effectively characterized using these methods.In the analysis of electronic target sequence,most of the existing methods are based on known data to speculate,or to analyze and predict a single target or a combination of several targets in the sequence.The former has very limited usage scenarios and is easily affected by the integrity of the known database.The latter cannot effectively use the overall topological characteristics of the sequence and has poor reliability.In view of the above problems,this thesis discusses and researches the modeling of electronic target sequence and the analysis of electronic target sequence based on the ideas of knowledge graph and graph neural network.Firstly,in terms of modeling,a hierarchical modeling scheme of electronic target sequence based on knowledge graph is proposed,and an electronic target sequence and formation data model is constructed,and a time-series electronic target sequence modeling scheme based on double-linked list is proposed based on the electronic target sequence model.At the same time,the 3-level indexing,storage and retrieval scheme of electronic target sequence is proposed based on NEO4 J database,and the encoding method of electronic target sequence is proposed based on hash function.By comparing with the existing modeling storage schemes,the feasibility and performance of the electronic target sequence modeling storage coding scheme in the simulation scenario are verified.Secondly,in terms of analysis,the task objectives of electronic target sequence analysis are discussed in this thesis.Based on task division,Ullmann’s algorithm and graph convolutional neural network algorithm are ussed to identify electronic target sequences in formation and task label association.Using breadth-first search algorithm and graph autoencoder,the hidden target node and hidden relationship edge estimation of electronic target sequence are carried out.At the same time,based on the temporal graph convolutional neural network,an improved temporal graph convolutional neural network is innovatively proposed in this thesis,and based on this network model,the structure prediction and parameter prediction of the time series electron target sequence are carried out.By constructing simulation scenarios and realistic data,the performance and practicability of the analysis algorithms in this thesis are verified and discussed,compared with the existing models.Finally,an electronic target sequence analysis and prediction system with integrated modeling and analysis research results is built,which provides reference and feasibility for the application of electronic target sequence modeling and analysis.