Font Size: a A A

Radar Imaging Techiniques Based On Sparse Bayesian Reconstruction Methods

Posted on:2016-04-07Degree:DoctorType:Dissertation
Country:ChinaCandidate:W G SuFull Text:PDF
GTID:1108330509461003Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Sparse Bayesian reconstruction method is one of the hot issues in the field of sparse signal processing, which not only considers the sparse of the signal, but also takes the priori statistical information of signal into account. Based on basic Bayesian theorem, by assigning a certain sparse signal promoting items to sparse prior distribution, the sparse Bayesian reconstruction method solves the sparse signal parameters by Bayesian inference. Sparse Bayesian reconstruction method is a parameter learning method, which automatically estimates the signal parameters in Bayesian inference process, and barely affected by human parameters intervention. In addition, taking the statistical information of noise into consideration, the sparse Bayesian reconstruction method can achieve better reconstruction performance at low SNR conditions. The radar imaging problem can be formulated as a sparse signal representation problem since the echo of the target can be represented by a few individual scatters at high frequencies. In view of the excellent reconstruction performance of the sparse Bayesian reconstruction method, this dissertation mainly discusses the novel radar imaging technology under the sparse Bayesian framework. The research of the dissertation mainly includes the following aspects.The first chapter describes the background and significance of the research. The current methods of sparse reconstruction are summarized and the advantages of sparse Bayesian reconstruction methods are pointed out. Subsequently, the status of the radar imaging method based on sparse Bayesian reconstruction method is overviewed, and the further study on this area in necessarity.The second chapter focuses on the sparse Bayesian learning(SBL) method in radar imaging. First, the sparse signal representation model and compressive sensing theory are introduced, and then the basic theory of SBL is in-depthly analyzed. Second, the sparse model of ISAR tomography imaging and the sparse model of the ballistic midcourse targets with micro-motion are established. The simulation results verify the effectiveness of the SBL based on these models. Subsequently, based on the ideal attributed scatterering center model, the SBL and other common sparse sparse reconstruction algorithms are used for SAR imaging. The influence of parameters on imaging performance is analyzed through experiments, which verifies the superiority of SBL. This chapter establishes the theoretical foundation to the sparse Bayesian reconstruction method used in radar imaging.Due to the shortcoming of poor timeliness and the sensitive to noise of the SBL, the third chapter firstly verifies that the expansion-compression variance-component based method(Ex Co V), which developed on the basis of SBL model, has a superior imaging performance in sparse ISAR Tomography imaging model. Secondly, on the one hand, under sparse linear FM ISAR imaging model, the ISAR imaging performance are analyzed and discussed based on the Ex Co V method; On the other hand, the frequency agile ISAR echoes are analyzed, and the disadvantages of existing imaging methods are pointed out. What’ more, the frequency agile sparse ISAR imaging model is established for the first time. By comparing with the existed imaging methods, the simulation results indicate the validity of the imaging programs based on Ex Co V method. Finally, based on the hybrid scattering center model, the sparse imaging model of SAR ground moving target imaging(GMTI) and SAR micro-move target imaging(MMTI) are established. Based on Ex Co V method, by comparing with the traditional imaging method, the simulation results are verified that the proposed imaging scheme can be more precise to estimate moving target parameters, and have better imaging performance in low SNR.Currently, the mostly radar imaging models based on sparse representation suppose that the radar observation model is very accurate. However, there are some phase errors in the radar echo actually because of the uncertainty of the radar imaging model. The fourth Chapter researches the radar imaging performance based on Ex Co V method when there are phase errors in radar imaging. Firstly, due to the introduce of phase errors, the Ex Co V-MLE algorithm is proposed to solve the sparse ISAR autofocus techniques. To reconstruct the sparse signals and estimate the phase errors, the Ex Co V-MLE algorithm is iteratively minimized the cost function with respect to the target scatterering coefficients and the phase error. Under different conditions and phase errors, by comparing the existing methods, the experiments results verify the validity of the algorithm; Secondly, since the rotation parameters of target is unknown in ISAR imaging, the joint estimation of the phase error and rotation parameters in ISAR imaging scheme is proposed based on Ex Co V method. Under different phase error conditions, the simulation results show the effectiveness of the imaging scheme. Finally, for the spotlight SAR model, the sparse imaging model is established under the existing of the observation position error. Then the SAR imaging algorithm combined with the estimation of the position error is proposed. In the first step, the scatterering coefficients are estimated using the Ex Co V method. In the second step, the conjugate gradient algorithm is used to estimate the observation position error. The proposed algorithm iterates the two steps several times until convergence. The simulation results verify that the proposed method can obtain high resolution spotlight SAR images under the random error conditions.Based on block sparse Bayesian reconstruction method, chapter 5 mainly studies the radar imaging with continuous scattering properties on block target in the imaging scene. Firstly, for the block sparse signal simulation, the block sparse Bayesian reconstruction algorithm named PCSBL(Patter-coupled Sparse Bayesian Learning) has a better reconstruction performance. Then range imaging based on block sparse Bayesian reconstruction methods is analyzed, the experiments proved the range image obtained by the PCSBL can better reflect the characteristics of the target. Secondly, the ISAR imaging based on ISAR tomography imaging model with different blocks structure of the target are simulated, the results show that the PCSBL algorithm still has a good imaging performance under this case. Finally, the sparse of strip SAR imaging model is established and the SAR imaging performance is explored based on block sparse Bayesian reconstruction methods. The simulated data are used to strip SAR imaging. Compared to the other block sparse reconstruction algorithm, the results show that the PCSBL algorithm can obtain more accurate SAR images with the strong of the structural complexity of target and increase of scatterering points.In Chapter six, the main innovations of the thesis are summarized. Drawbacks are pointed out and possible further research tracks are discussed.
Keywords/Search Tags:Radar imaging, ISAR, SAR, sparse representation, compressed Sensing, prior information, sparse Bayesian reconstruction, SBL, Bayesian inference, expectation maximization, ExCoV, phase error, conjugate gradient, block sparse Bayesian reconstruction model
PDF Full Text Request
Related items