Ballistic Missile Micro-doppler Feature Extraction Based On Radar Measurements

Mid-course phase is the one that is most attractive for interception of a ballistic mis-sile (BM) warhead. However decoy releasing makes this phase of the flight the mostdifficult for target discrimination. Extracting appropriate feature parameters from radarreturns is a key technique in radar target discrimination of a BM. The aim of this paperis to analyze the micro-Doppler features of ballistic missiles during flight, and investigatemethods to extract the appropriate feature parameters.InChapterOne, thedevelopmentstatusofthetargetdiscriminationtechnologyoftheBM is reviewed firstly. And then the characteristics and challenges of the technology inthe mid-course phase are briefly introduced. The main methods of the feature extractionand target discrimination of the BM in the mid-course phase are summarized as well. Themajor contributions of this dissertation are listed at the end of this chapter.In Chapter Two, the geometric model of the BM observing by a radar is establishedsystematically. The shield effect of the spinning BM and the dynamic RCS scintillationare deduced in detail. The micro-Doppler effect of the radar echo from a BM with micro-motion is investigated, and the relationship between the the RCS sequence period and theprecession frequency of the BM is discussed. The study in this chapter is the foundationof the following research on the feature extraction of the BM targets.In Chapter Three, the research mainly focuses on the precession frequency estima-tion based on the BM dynamic radar cross section (RCS) sequence. Firstly the influencesof the radar measure system to the RCS distribution is derived. The scintillation effectis taken into account and modeled as a multiplicative noise. And then the dynamic RCSdistribution model of a precessing conical missile is established. Two pseudo maximumlikelihoodestimation(MLE)approachestoextracttheparameterofmissileprecessionfre-quency are proposed. The first approach ignores the additive noise (i.e., assuming the in-finite signal-to-noise ratio (SNR)). The second approach enforces a Gaussian distributionon both additive and multiplicative noise components. The Cram′er-Rao Lower Bound(CRLB) corresponding to the maximum SNR scenario is derived. Simulations indicatethat accounting for the multiplicative noise in the estimation significantly improves esti-mationperformance,andalsoshowtherobustnessandvalidationoftheproposedmethods. In Chapter Four, a method to estimated the micro-Doppler modulation frequencythroughanalyzingtheperiodicstructureoftheresultingtime-frequencydistribution(TFD)spectrogram is presented. The relationship between the texture period of the TFD ofa spinning BM and the micro-Doppler modulation frequency is analyzed from the phe-nomenology perspective. Based on the texture analysis in the image processing theory,the resultant TFD spectrogram is treated as an image (with periodic features along thetime axis). The2-D DFT (followed by autocorrelation) is then used to exploit the wellknown periodicity of the micro-Doppler signature in the TF domain, to better estimate themicro-Doppler modulation frequency via a simple cost effective and system friendly way.The method presented in this chapter has low computational cost, and does not need anyprior information about the signals, which satisfies the special demands of a BM targetrecognition system. The simulation results also indicate that the method work well underthe conditions of low signal to noise ratio (SNR).In Chapter Five, aiming at saving the time resources in the ballistic missile defense(BMD) system and taking full advantage of radar measurements, a feature estimationmethod through a set of non-uniformly sampled radar signal is investigated. The esti-matedfeaturesincludethetheprecessionfrequencyandthescatteringcenters’distributioninformation of a BM with micro-motion. According to the sparsity nature of the BM scat-tering centers in the high frequency region, the parametric dictionary based on the sparsecomponent analysis (SCA) theory is constructed and the radar signal sampling model isestablished. And then, the algorithms of the nonlinear least square (NLLS) and the or-thogonal matching pursuit (OMP) are jointly employed to reconstruct the radar signal.Therefore, the precession frequency and the scattering centers’ location of the precess-ing BM are estimated according to the sparse representation of the signal. The methodproposes a new notion for the feature extraction of the BM target under the restriction oflimited time resource in the ballistic missile defense system.In Chapter Six, the work of the dissertation is summarized and the related furtherresearch is discussed.