STAP Technique For Airborne Radar Based On Sparse Representation

Clutter suppression plays a crucial role for airborne radar to accomplish its multiple objectives,and space-time adaptive processing(STAP)is proven to be an effective way for suppressing the clutter received by airborne radar.For traditional STAP approaches,the number of independent and identically distributed(IID)training samples should be no smaller than twice the system degrees of freedom to yield optimal performance.However,in practice,the clutter environment encountered by airborne radar tends to be non-homogeneous and non-stationary,thus it may be impossible to acquire sufficient IID training samples.The system performance will degrade drastically due to insufficient IID training samples.Aiming at improving the clutter suppression performance of airborne phased array radar under the situation of finite-training-sample,this dissertation studies the STAP theory and algorithms based on sparse representation techniques.In chapter 1,the background and significance of this dissertation are illustrated.The development history of STAP technology is elaborated,and the state of the art in STAP approaches is summarized.Some discussions on sparse representation STAP technique are also made.In chapter 2,the basic principles in STAP technology are introduced.The signal model utilized in sparse representation and the popular sparse representation approaches are described.After a review of the construction of STAP dictionary,an investigation on the sparsity of clutter signal in STAP radar is presented.Then,the possible approaches and advantages of sparse representation STAP technique are discussed.Chapter 3 studies the sparse representation STAP technique from the viewpoint of clutter whitening.And two algorithms that can effectively estimate the clutter covariance matrix in low training support situation are developed.On one hand,a sparse representation STAP algorithm is developed based on the subspace-augmented multiple signal classification theory.This algorithm can effectively estimate the clutter covariance with only few training samples and easy parameter settings.On the other hand,to counteract the slow convergence of sparse Bayesian learning algorithm,a fast converging sparse Bayesian learning algorithm is proposed.Then,the proposed algorithm is utilized to estimate the clutter covariance matrix in low training support situation.The performance of clutter whitening STAP technique in non-homogeneous clutter environment is greatly enhanced by the algorithms proposed in this chapter.Chapter 4 studies the sparse representation STAP technique from the perspective of clutter nulling.And two algorithms that can effectively estimate the clutter subspace in low training support situation are proposed.On one hand,enlightened by the idea of subspace augmentation,the entire clutter subspace is divided into the direct portion and the supplemented portion.The direct portion is estimated from the limited training samples,while the supplemented portion is constructed by some space-time steering vectors selected from the STAP dictionary.Then,the entire clutter subspace is estimated by augmenting the direct portion with the supplemented portion.This algorithm can effectively estimate the clutter subspace with limited training samples.On the other hand,since the training samples are inevitably contaminated by noise,it is impossible to estimate a pure clutter subspace from training samples.To cope with this problem,an algorithm which can estimate a pure clutter subspace is proposed based on an atom selection criterion which is particularly designed for the problem.In this algorithm,the clutter subspace is spanned by a group of atoms selected from the STAP dictionary.The performance of clutter nulling STAP technique in non-homogeneous clutter environment is greatly enhanced by the algorithms proposed in this chapter.Chapter 5 focuses on the suppression of non-stationary clutter in range ambiguous condition.With the assumption of using a rectangular array,a sparse representation based subspace projection preprocessing algorithm is derived.In this algorithm,the clutter in the azimuth-elevation domain is divided into two categories: short-range and long-range.And the clutter subspace corresponding to the short-range clutter is estimated by exploiting sparse representation technique.Then,the short-range clutter can be eliminated by subspace projection processing.After that,the rectangular array data is synthesized into a linear array data by conventional beamforming,and the classical STAP methods can be applied to the data to suppress the residual clutter.The proposed algorithm does not require the geometrical information between the ground and the radar array,nor the parameters of the platform.It can eliminate the short-rang clutter separately for each range cell using the data from the same range cell.Chapter 6 studies the off-grid problem encountered when constructing the STAP dictionary.And a clutter nulling STAP algorithm is designed based on the idea of local searching.By introducing the local searching step into the algorithm,a more accurate clutter subspace can be estimated,and thus the clutter suppression performance can be enhanced.This algorithm can effectively alleviate the performance degradation of sparse representation STAP technique caused by off-grid problem.Chapter 7 summarizes the dissertation,and some perspectives about the future work are proposed.