Radar Threat Analysis And Recognition Based On Machine Learning Techniques

This thesis researches for the analysis and recognition method of radar threat signal based on machine learning.The rapid development of radar technology,modern radar parameters are flexible and generally has the characteristics such as multi-mode,adaptive agility,etc.Variable parameters bring more difficulty to the analysis and identification of radar transmit signals,and the effective analysis and identification of radar signals is the key to dealing with radar threats.There are two possible reasons for the changes of radar parameters.One is the modulation of radar parameters according to its own system and mission requirements.The other is that radar adjusts the operating parameters according to the external environment and the characteristics of its target in order to optimize the working effect.The former is the necessary parameter change rules of the radar under different tasks or different systems,the latter is the adaptive adjustment behavior of the radar affected by the environment.In this thesis,the analysis and identification of radar threat signals are based on the above two perspectives,research for the classification and recognition methods of multi-mode radar and the identification methods of radar adaptive behavior characteristics.The main research contents include the following:(1)Perform necessary research and analysis on radar objects.Firstly,construct a hierarchical radar signal pattern description method based on radar parameter variation mode and parameter value division.This method can effectively classify and identify radar signals with multi-parameter modulation rules.In addition,define the concept of radar adaptive waveform behavior.The adaptive waveform behavior of radar is defined as that radar changes its own transmit waveform parameters according to the external environment.Introduce the basic principle of radar waveform selection,research for the algorithm of radar tracking adaptive waveform optimization and verify the effect of waveform optimization by simulation.(2)Study and build a model to recognize radar signal pattern based on convolutional neural network and density clustering algorithm.This model takes the sequence of radar pulse parameters as input,including one identification and two classifications.First,the convolutional neural network is used to identify the parameter sequence to obtain the change pattern of the parameters such as stagger,slip,dwell and switch,etc.The first classification of radar signals is according to the recognition results of the change pattern of parameters.On the basis of the first classification,the clustering algorithm performs the second classification according to the value characteristics of each parameter,and finally divides the radar signal into different modes.Construct the radar signal frame to generate experimental data,simulate to verify the various parts of algorithm and overall algorithm.This method can classify and identify the radar threat signal effectively,and the characteristics of classified signal can be used to judge the actual working mode of the radar to determine the threat level of the radar.(3)Research for the method to recognize adaptive waveform behavior of radar on the base of the radar adaptive waveform behavior concept and radar adaptive waveform selection principle.In this thesis,the radar adaptive waveform optimization process is abstracted into a decision-making system which takes the target characteristics,clutter environment,interference and the transmitted waveform parameters at previous moment as input,and the transmitted waveform parameters at next time as output.Thereby,the radar adaptive waveform behavior recognition is transformed into the system identification of this system.On this basis,this thesis researches for the system identification method of radar adaptive waveform selection system through neural network algorithm.The neural network is trained to predict the waveform parameters of radar under the influence of external environment.The simulation results show that the proposed method can effectively learn the adaptive behavior of radar waveform.This method realizes to predict the operating parameters of radar which has adaptive waveform selection,according to target characteristics and interference,so as to deal with radar threats effectively.