Research On Airborne Radar STAP Method Based On Knowledge-aided

The primary task of airborne radar is to detect the target,but the echo signal received by the radar not only contains the target signal,but also inevitably includes some clutter signals in the scene at the same time,and the clutter signals seriously affect the target detection performance of the radar.Therefore,the prerequisite for effective target detection is the need for clutter suppression of the echo signal.Space-time adaptive processing(STAP)is a typical clutter suppression technique,but the difficulties in obtaining a sufficient number of samples to estimate the covariance matrix in a non-uniform environment have severely degraded the performance of the STAP algorithm.For the problem of insufficient samples,the knowledge-aided STAP(KA-STAP)technique is proven to be an effective solution,so this paper focuses on the KA-STAP technique in airborne radar,with the main research content as follows:(1)The KA-based Persymmetric structured covariance matrix estimation is studied.Firstly,the Persymmetric structured covariance matrix shrinkage estimation method is studied under Euclidean distance based on the clutter prior knowledge and Persymmetric property,and the simulation implementation and performance analysis of the method are carried out.Secondly,for the problem that the optimal solution expression of the above method cannot evaluate the gap between the prior covariance matrix and the true covariance matrix,the whitening capability-based dimensionality reduction STAP method for airborne radar is studied.A Persymmetric-based pre-whitening performance evaluation algorithm for the cell under test(CUT)is proposed using prior knowledge and Persymmetric covariance structure,and based on the whitening capability of the ideal covariance matrix.The algorithm is robust to prior knowledge of different accuracy and is able to maximize the whitening of the observed disturbance data.At the same time,considering the huge computational effort caused by high-dimensional STAP,the extended factored approach(EFA)is used to optimize the algorithm for dimensionality reduction processing.The simulation results show that the algorithm can effectively improve the signal-to-interference-plus-noise ratio(SINR)and enhance the STAP performance even in the presence of certain errors in prior knowledge.(2)The structured covariance matrix estimation based on geometric methods is studied.Since the algorithms in(1)rely on the prior covariance matrix constructed using heterodyne prior knowledge,they are not applicable to the case when no prior knowledge is available.For this case,following the geometric paradigm,two structured covariance matrix estimation algorithms are proposed.Specifically,starting from a set of training data,two structured sample covariance matrices are constructed by considering different covariance matrix structure information(i.e.,Persymmetric or Toeplitz structures).By exploiting the characteristics of positive-definite matrix space and imposing an upper bound constraint on the condition number,the two minimization problems are established with the Frobenius norm between the ideal covariance matrix and the corresponding structured sample covariance matrix as the objective functions.The problems are transformed and the closed-form solutions of the estimators are finally obtained.The simulation results show that in both cases,the proposed two algorithms can estimate the covariance matrix more accurately compared with the reference algorithm,thereby effectively improving the SINR and enhancing the radar interference suppression performance,and the optimization effect is more significant in the case of small samples.