The spread of sea clutter in the Doppler domain caused by multi-mode propagation would make it difficult for over-the-horizon radar to detect low observable group targets. To solve this problem, a multi-mode clutter suppression method based on space–frequency cascaded adaptive processing is proposed by utilizing the spatial diversity characteristics of multiple-input–multiple-output radar. Then, the low observable group targets can be detected. The proposed method adopts the idea of space–frequency stepwise processing. First, the multi-stage Wiener filter ESPRIT algorithm is exploited to estimate the direction of departure and direction of arrival of different path signals, and then the multi-mode clutter is adaptively transformed into single-mode clutter based on the estimated angles. Second, the atomic norm minimization algorithm and Vandermonde decomposition are used to estimate the steering vector of the first order Bragg frequency of sea clutter, whose path contains the group targets. Finally, the least squares algorithm is used to realize the adaptive suppression of sea clutter in the frequency domain. Because there is no need to estimate the signal covariance matrix and its eigenvalue decomposition and conduct spectral peak searching, the proposed method greatly reduces the computation of the angle estimation. Compared with the existing second-order blind identification and spatial smoothing SOBI algorithms, the proposed algorithm does not need the information of the first order Bragg frequency of sea clutter, so it can improve the output signal-to-clutter-noise ratio when there is an error in the first order Bragg frequency of sea clutter. Therefore, the proposed algorithm is more conducive to the detection of low observable group targets. Simulation results are given to demonstrate the effectiveness of the proposed method.

由多模式传播引起的海杂波在多普勒域中的扩散将使超视距雷达难以检测到低可观测的目标群。针对这一问题，利用多输入多输出雷达的空间分集特性，提出了一种基于空频级联自适应处理的多模式杂波抑制方法。然后，可以检测到低可观察到的组目标。所提出的方法采用了空间频率逐步处理的思想。首先，利用多级维纳滤波器ESPRIT算法估计不同路径信号的离开方向和到达方向，然后根据估计的角度将多模杂波自适应地转换为单模杂波。第二，运用原子范数最小化算法和范德蒙德分解估计海杂波一阶布拉格频率的转向矢量，该路径的路径包含群目标。最后，采用最小二乘算法实现频域中海杂波的自适应抑制。由于不需要估计信号协方差矩阵及其特征值分解并进行频谱峰值搜索，因此该方法大大减少了角度估计的计算。与现有的二阶盲识别和空间平滑SOBI算法相比，该算法不需要海杂波的一阶布拉格频率信息。因此，当海杂波的一阶布拉格频率出现误差时，它可以提高输出信噪比。因此，所提出的算法更有利于低可观测目标的检测。仿真结果证明了该方法的有效性。