To improve the performance of the recently developed parameter‐dependent sparse recovery (SR) space–time adaptive processing (STAP) algorithms in real‐world applications, the authors propose a novel clutter suppression algorithm with multiple measurement vectors (MMVs) using sparse Bayesian learning (SBL) strategy. First, the necessary and sufficient condition for uniqueness of sparse solutions to the SR STAP with MMV is derived. Then the SBL STAP algorithm in MMV case is introduced, and the process for hyperparameters estimation via expectation maximisation is given. Finally, a computational complexity comparison with the existing algorithms and an analysis of the proposed algorithm are conducted. Results with both simulated and the Mountain‐Top data demonstrate the fast convergence and good performance of the proposed algorithm.

In this study, a novel method for overcomplete dictionary construction for sparse recovery (SR) space–time adaptive processing (STAP) is proposed, which can effectively mitigate the off‐grid effect. The proposed method utilises the clutter ridge, calculated via some prior knowledge such as radar system parameters and modern inertial navigation system information, to discretise the grids in the angle–Doppler plane. The authors also investigate the accuracy of clutter ridge in the presence of several typical non‐ideal factors such as array errors, intrinsic clutter motion and aircraft crab. In particular, a parameter is introduced to adaptively adjust the grid interval to avoid the strong column coherence of the dictionary. The simulation results show that the signal‐to‐clutter‐plus‐noise ratio for the SR STAP with the proposed dictionary is significantly improved, compared with the one utilising conventional dictionary. Moreover, it is shown that the novel method not only works more robustly for the side‐looking array, but also has better performance for other array orientations.

The space–time adaptive processing (STAP) technique has achieved good clutter suppression performance for a side-looking airborne radar; however, it suffers from severe performance degradation in nonside-looking airborne radar. This is because the clutter distribution varies considerably with range. The 3D STAP can provide better performance compared with the traditional STAP methods in such a nonstationary clutter environment, but it requires high computational complexity and sample support. In this letter, the characteristic of azimuth–elevation–Doppler 3-D beam pattern for the planar array is explored, and a novel reduced-dimension scheme combined with this characteristic is proposed. We further developed three basic reduced-dimension structures according to this scheme. Furthermore, we also prove that the first and third structures are just the direct extension of the traditional 2-D generalized multibeam and joint-domain localized methods. The second one is an original structure that performs local reduced-dimension processing with three orthogonal planar windows. Numerical results verify that the proposed structures have significant advantages in reducing computational load and training sample numbers compared with existing 3D STAP methods. In particular, the second structure shows the best comprehensive performance in the bad samples environment.

Space‐time adaptive processing (STAP) methods can provide good clutter suppression potential in airborne radar systems. However, the performance of these methods is limited by the training samples' support in practical applications. To address this issue, a deep learning framework for STAP is developed. First, the clutter space‐time data and their exact clutter covariance matrices (CCMs) are simultaneously modelled via simulation, in which various non‐ideal factors such as aircraft crabbing, array errors, and internal clutter motion with all possible levels in practice are all considered. Then, a multi‐layer two‐dimensional convolutional neural network (CNN) is developed. In this CNN, low‐resolution angle‐Doppler profiles estimated by a few training samples are used for the input and the high‐resolution counterpart obtained by the exact CCMs are utilized for the labels. Once trained, the CNN can be used to predict the high‐resolution angle‐Doppler profile using a few measured data in near real time. The high‐resolution clutter spectrum can be further calculated using the space‐time steering dictionary and the above obtained profile. Finally, the CCM of the measured data can be constructed and the space‐time weight vector can also be achieved. Compared with recently developed sparsity‐based STAP methods, the performance of the proposed method is better and the computational load of it is far fewer, and therefore more suitable for real‐world implementation. The simulation results have demonstrated the superiority of the proposed method in both clutter suppression performance and computation efficiency.