Multiple-input-multiple-output (MIMO) synthetic aperture radar (SAR) (MIMO-SAR) has numerous potential applications in fields, such as medical diagnosis and security screening. In near-field MIMO-SAR imaging, the virtual array is typically defined as the midpoint of the transceiver array. Based on the virtual array, the multistatic data can be converted to monostatic data for imaging. However, these approaches suffer from low accuracy and cannot reconstruct large scenes, despite their small computational load. In this article, we propose a novel near-field MIMO-SAR imaging algorithm based on range compensation. The convolution between transmitting and receiving elements of the MIMO array defines the positions of virtual array's elements, with the center of the small scene selected as the reference point. The range difference between the transceiver array and the corresponding virtual array to the reference point is calculated. An approximate expression of the range difference is also derived, which can effectively reduce the computational load associated with the range compensation. The conditions for partitioning large scenes are derived by analyzing the range difference between the reference point and other scattered points with respect to the antenna elements. Based on the conditions, the large scenes are divided into blocks. The final image is obtained by a weighted sum of each block scene image. Simulation and experimental results on the designed MIMO radar near-field imaging system demonstrate that the proposed algorithm can effectively reconstruct high-resolution scene images of arbitrary size.

Neural networks for synthetic aperture radar (SAR) automatic target recognition often encounter overfitting challenges owing to limited training samples. Moreover, the azimuth angle of SAR, a vital parameter for improving network generalization, is frequently disregarded in most models. In response, we propose MIGA-Net, a classification neural network that effectively perceives azimuthal information using multi-view images to improve classification performance. Specifically, we quantize low-dimensional azimuthal values for sample-limited scenarios. Then, we utilize encoded image sequences as training data because they encompass spatial context information compared to individual images. After extracting features of the sequence samples through convolutional layers, we design a two-layer output module. One layer converts these sequence features into graph data. Then the dense graph attention network (GAT) extracts contextual features from the graph data for angle estimation. Simultaneously, another layer combines these features for target classification. During the network training, the GAT module can extract image azimuth features with powerful information aggregation capabilities. It supervises the convolutional layers to learn azimuth features, which are fused with class features from another layer to obtain a more structured feature domain. This feature domain significantly enhances the classification performance of the network. Experiments conducted on the moving and stationary target acquisition and recognition (MSTAR) dataset have proven the superior performance of the proposed method, achieving at least 1% higher accuracy compared to other state-of-the-art algorithms.

Combining inverse projection and vote accumulation methods, we propose a novel 3D geometric reconstruction method based on inverse synthetic aperture radar (ISAR) image sequence for triaxially stabilized space targets. Considering that the translation of a triaxially stabilized space target results in a large angular rotation of the radar line of sight (LOS) and an attitude adjustment rotation of the target itself, the projection relationship between radar imaging plane and a triaxially stabilized space target is constructed first. Subsequently, we perform an inverse projection on the dominant points extracted from each imaging plane to determine the location of the 3D candidate scatterers and assign a voting value of 1 to each position. By accumulating these vote values, we can obtain the total number of votes for the candidate scatterer position. We treat candidate 3D scatterers with a large cumulative number of votes as actual scatterers. Thus, the target 3D geometry can be obtained. Compared to traditional matrix factorization based 3D reconstruction methods, this method is more accurate and ensures that the target pose is consistent with the actual pose. At the same time, compared with 3D reconstruction methods based on energy accumulation ideas, this method produces fewer false scatterers and has higher credibility. Experimental results based on the simulated point target and electromagnetic data are presented to validate the effectiveness and robustness of the proposed method.

Due to the limited information of two-dimensional (2D) radar images, the study of three-dimensional (3D) radar image reconstruction has received significant attention. However, the target attitude obtained by the existing 3D reconstruction methods is unknown. In addition, using a single perspective, one can only get 3D reconstruction result of a simple target. For a complex target, due to occlusion and scattering characteristics, 3D reconstruction information obtained from a single perspective is limited. To tackle the above two problems, this paper proposes a new method for multi-perspective 3D reconstruction and single perspective instantaneous target attitude estimation. This method consists of three steps. First, the result of 3D reconstruction with unknown attitude is obtained by the traditional matrix factorization method. Then, in order to obtain the attitude of a target 3D reconstruction, additional constraints are added to the projection vectors which are computed from the matrix factorization method. Finally, the information from different perspectives are merged into a single layer information according to certain rules. After the information fusion, a multi-perspective 3D reconstruction structure with better visibility and more information is obtained. Simulation results have proved the effectiveness and robustness of the proposed method.

Two-dimensional (2D) large-plane multiple-input-multiple-output (MIMO) array antennas with wide coverage can obtain high-resolution three-dimensional (3D) images, which have great potential for applications in security surveillance, medical diagnostics, and nondestructive evaluation. Cross MIMO arrays with low hardware cost and complexity are one of the effective choices for 2D large-plane MIMO arrays. However, the current imaging algorithms corresponding to the cross MIMO arrays are computationally intensive. In this paper, a fast wavenumber domain 3D near-field imaging algorithm for cross MIMO array is proposed. According to signal characteristics of the cross MIMO array, the signal is effectively expressed in the convolution form. The convolution signal is converted to wavenumber domain to achieve horizontal and pitch direction imaging, and wavenumber domain operation can effectively improve computational efficiency. Meanwhile, this algorithm uses distance approximation reasonably according to the distance characteristics of the actual scene, greatly improving the computational efficiency. Simulation and experimental results show that, featuring faster computational efficiency, the proposed algorithm can obtain images that have the similar quality as the images obtained by the commonly used algorithm.