Study On High-Resolution Sparse Subband ISAR Imaging

Although inverse synthetic aperture radar(ISAR)can provide relatively high range resolution,small targets such as space debris and microsatellites only occupy several or even one range unit.In this case,it is necessary to further increase the range resolution to achieve a fine description of target characteristics,such as size,shape,orientation,and motion.In order to reduce the hardware cost,we could deploy multiple ISAR to observe the target in different frequency bands,and achieve sparse subband high-resolution imaging by advanced signal processing algorithms.This thesis focuses on the key problems and difficulties in sparse subband ISAR imaging.First,the signal model of sparse subband ISAR imaging is established.Then,the high-resolution sparse subband ISAR imaging algorithm and coherent processing based on spectral estimation are studied.To deal with complex observation conditions such as low signal-to-noise ratio,imaging methods based on probabilistic modeling and Bayesian inference are studied in detail.Finally,the graphical user interface(GUI)of the revelant algorithms is developed.The research work of this thesis will provide theoretical and technical support for improving the capability of space target detection of available ground-based imaging radar.The content of this thesis includes the following four parts.In the first part,we briefly introduce the sparse subband observation model and the all-pole signal model,and then study sparse subband high-resolution imaging algorithm based on multiple signal classification(MUSIC)and its improved version.On this basis,we propose a coherent processing algorithm for multiple subband signals based on spectral analysis.The effectiveness of the algorithm is verified by simulated data.To tackle the problem that it is difficult for the spectral estimation method to accurately estimate the model order in low signal-to-noise ratio,the second part studies sparse subband ISAR imaging based on Bayesian inference.Firstly,the sparse signal representation theory is reviewed and the problem of sparse subband imaging is transformed to a sparse coefficient vector estimation problem.With this model,the Bayesian probabilistic model is constructed according to the prior information of both scatterer distribution and noise.Then,the expectation-maximization(EM)algorithm and the fast relevance vector machine(f RVM)algorithm are applied to solve the model,respectively,to reconstruct the full-band echoes and achieve well-focused two-dimensional ISAR imaging.The effectiveness of the algorithms is verified by simulated data.In the third part,we systematically study high-resolution sparse subband ISAR imaging based on Bayesian hierachical priors.Considering the limitations of Gaussian priors in data description and sparse-promotion,a probabilistic graphical model based on the Gamma-Gaussian hierarchical prior is proposed,which is then solved effectively by variational inference(VI).To improve the estimation accuracy,a hierarchical dictionary is further designed.Combined with the proposed sparse subband imaging method,an algorithm for multi-band coherent processing based on the minimum entropy criterion is introduced.To solve the problem of inaccurate estimation when the scatterers deviate from the dictionary grid,an off-grid variational sparse Bayesian learning algorithm is proposed based on a modified Bayesian graphical model,which obtains well-focused two-dimensional imaging with deviated scatterers.Finally,the effectiveness of the algorithm is verified by simulated data.In the fourth part,we write the graphical user interface(GUI)for revelant sparse subband imaging algorithms.The composition structure,basic design principles and procedures of MATLAB GUI are reviewed,and then a GUI interface for sparse subband ISAR imaging is designed by applying the GUIDE development tool.Such interface is capable of realizing sparse subband high-resolution imaging of airplanes with various categories.