Information Theory Based HRRP Target Recognition

In the area of automatic target recognition(ATR),high resolution range profile(HRRP)contains the information of target structure and is easy to be acquired,and therefore makes HRRP attracted many attentions in the society of ATR.Since Shannon created the modern information theory,which can improve the effectiveness and reliability of the system,it has been gradually deepening into the area of ATR.Existing studies have shown that facing the increasing complex and digitizing military operations,using the concepts of information theory can improve performance of radar target recognition effectively.This thesis studies the relative theory and technology of information theory for HRRP signal processing and target recognition.The main content is summarized as following:First,considering the HRRP signal obtained by heterogeneous radar sense network(HRSN),this thesis deploys distributed compressed sensing(DCS)in of HRRP signal and proposes three HRSN scenarios for the first time.We apply the joint sparsity mode I(JSM-1)of DCS into the HRSN-obtain HRRP signals.Firstly compressing the HRRP signal in every single sensor independently,the scheme then reconstructs the signal at the fusion center jointly.Aiming at the different capabilities among the sensors in HRSN,this thesis proposes following HRSN scenarios: different types of measurement matrices and same numbers of measurements,the same types of measurement matrices and different numbers of measurements,different types of measurement matrices and different numbers of measurements.In the end,Monte Carlo experiments are conducted to simulate and analyze these three scenarios.The results show that joint decoding using all-Fourier measurement matrices performs better than that of all-Gaussian measurement matrices.The performance of using different types of measurement matrices is between above of two.Second,this thesis proposes the minimum resistor-average(MRA)distance criterion in the multiples HRRPs recognition problem.Traditionally HRRP target recognition is based on single profile and neglects a sequence of independent HRRPs obtained in many radar systems.This method suffers the problem of HRRP's target aspect sensitivity,which will lead to poor recognition performance.The proposed algorithm takes advantage of multiples HRRPs.At first,the average range profile and the variance profile are extracted together as the feature vectors for both training data and test data representation.Then MRA distance criterion is used to determined the class of the target.Monte Carlo experiments show that comparing with the performance of maximum correlation coefficient template matching method(MCC-TMM)and minimum Kullback-Leibler(MKL)distance criterion,the proposed method can improve the ratio of correct recognition.In addition,MRA,MCC-TMM and MKL are also used in radar sensor networks.Experimental results also show that the MRA distance criterion offers the improvement of correct recognition ratio.Finally,facing the problem that when applying Dempster–Shafer(D-S)evidence theory in HRRP target recognition,the evidence conflict problem will be encountered,this thesis proposes an information-theory based combination rule.Under the circumstance of the interference of radar echo signals and the evidence conflict problem met in fusing results of several HRRP recognition algorithm,this thesis compares the results of using the original D–S rules,Yager combination rules,Murphy evidence-average combination rules,the evidence-distance weighted combination rules and the proposed combination rules.The proposed method considers both relationships and inner properties of evidences,which conforms to reality.The investigation of this thesis expands information theory in HRRP target recognition,and provides an efficient method for target recognition in modern complex battlefield environment.