High Resolution ISAR Imaging Based On Sparse Bayesian Learning

High resolution inverse synthetic aperture radar(ISAR)images of air and space targets contain important information such as shape and size,and the imaging quality directly affects the accuracy of subsequent feature extraction and automatic target recognition process.In practice,ISAR imaging is often confronted with low signal-to-noise ratio,data missing,random phase error and target maneuvering,which will seriously degrade the performance of traditional imaging methods based on Fourier transform.Because ISAR imaging based on sparse Bayesian learning can make full use of the statistical information of target scatterers and the environment,it provides a new way to imaging under complex observation conditions discussed above.This thesis focuses on the research of high resolution ISAR imaging method based on sparse Bayesian learning,which mainly includes sparse modeling and probabilistic graphical modeling;high resolution imaging of uniformly rotating targets based on sparse Bayesian learning;and high resolution imaging of maneuvering targets based on sparse Bayesian learning.The relevant studies have fully explioted the sparsity nature of the ISAR image,proposed high resolution ISAR imaging technqiues based on advanced machine learning theory such as sparse Bayesian learning,and achieved well-focused high resolution imaging of uniformly rotating and maneuvering targets under complex observation conditions.The research will provide theoretical and technical supports for improving the detection and imaging capability of imaging radars in space situation awareness.This thesis mainly consists of three parts:In the first part,high resolution ISAR imaging based on Gamma-Gaussian prior is studied.Firstly,the sparse observation model is established.Then,the probabilistic graphical model is constructed by imposing the Gamma-Gaussian prior to the distribution of scatterers,based on which the model parameters are estimated by expectation maximization(EM)and variational Bayesian inference(VBI),respectively.After that,a high resolution imaging and random phase error correction method based on VBI and maximum likelihood(ML)estimation is proposed.Finally,the effectiveness of the proposed method is verified by Monte Carlo experiment and measured data processing.The second part studies high resolution ISAR imaging based on the spike-and-slab prior.In order to further improve the flexibility of data description and improve the accuracy of sparse reconstruction,a probabilistic graphical model is constructed by imposing the spike-and-slab prior to the distribution of scatterers.Then,expectation propagation(EP)is proposed to realize sparse singal reconstruction.On this basis,a high resolution imaging and random phase error correction method based on EP and ML estimation is proposed.Finally,the effectiveness of the proposed method is verified by Monte Carlo experiment and measured data processing.The third part studies ISAR imaging of maneuvering targets in low SNR based on sparse Bayesian learning.Firstly,a sparse observation model of the maneuvering target is established,where the parametric over-complete dictionary is constructed by taking the rotation speed and rotation acceleration as parameters.Then,the Gamma-Gaussian prior is imposed on the distribution of scatterers and the probabilistic graphical model is constructed.Finally,the high resolution imaging and dictionary parameter estimation method based on VBI and ML are proposed,and the effectiveness of the method is proved by simulated and measured data.