Transformer-based networks, which can well model global characteristics of inputted data using attention mechanism, have been widely applied to hyperspectral image (HSI) classification and achieved promising results. However, existing networks fail to explore complex local land cover structures in different scales of shapes in hyperspectral remote sensing images. Therefore, a novel network named multi-scale and cross-level attention learning (MCAL) network is proposed to fully explore both global and local multi-scale features of pixels for classification. In order to encounter local spatial context of pixels in transformer, a multi-scale feature extraction module (MSFE) is constructed and implemented into transformer-based networks. Moreover, a cross-level feature fusion module (CLFF) is proposed to adaptively fuse features from hierarchical structure of MSFEs using attention mechanism. Finally, the spectral attention module is implemented prior to the hierarchical structure of MSFEs, by which both spatial context and spectral information are jointly emphasized for hyperspectral classification. Experiments over several benchmark datasets demonstrate that the proposed MCAL obviously outperforms both CNN-based and transformer-based state-of-the-art networks for hyperspectral classification.

Conventional synthetic aperture radar (SAR) ground moving target detection is only effective for targets with high radar cross section (RCS), and its performance is degraded by focusing distortion caused by the motion of the target. This letter proposes a new low-RCS moving target detection method which exploits the target’s shadow using high-resolution sequential images captured by a Ka-band video SAR system. First, the characteristics of target shadow are investigated, which are mainly influenced by the size of the target, the incidence of the radar beam, and the target velocity. Then the presented method based on video SAR system is described, and its moving target detection performance is also analyzed. As a preliminary study, a simulated experiment is conducted in which the signal of a low-RCS moving target is simulated and inserted into real airborne Ka-band video SAR images. The results demonstrate the feasibility of the proposed method for low-RCS moving target detection.

The azimuth multichannel synthetic aperture radar (SAR) system is used in ground moving target indication (GMTI). But for Ka-band SAR, the conflicts between high gain and wide swath coverage are very strict. The digital beamforming (DBF) scan-on-receive (SCORE) algorithm can solve this problem very well. This letter introduces a multichannel SAR system with three receiving antennas in the azimuth direction. For each azimuth antenna, eight subapertures are arranged in the elevation direction in order to obtain higher gain and a lower data rate through the DBF-SCORE algorithm. Airborne flight experiments with this system are accomplished successfully and quantities of experimental data are acquired. In this letter, we first derive the signal forms and weighting coefficients in the DBF-SCORE procedures. Then, different GMTI methods including along-track interferometry and displaced phase center antenna are introduced for azimuth multichannel SAR-GMTI based on the DBF-SCORE technology. Finally, airborne raw data with cooperating moving targets are processed successfully, and the estimation errors of the radial velocities meet the expected accuracy requirements.

Due to a unique imaging mechanism, Synthetic Aperture Radar (SAR) images typically exhibit degradation phenomena. To enhance image quality and support real-time on-board processing capabilities, we propose a lightweight deep generative network framework, namely, the Lightweight Super-Resolution Generative Adversarial Network (LSRGAN). This method introduces Depthwise Separable Convolution (DSConv) in residual blocks to compress the original Generative Adversarial Network (GAN) and uses the SeLU activation function to construct a lightweight residual module (LRM) suitable for SAR image characteristics. Furthermore, we combine the LRM with an optimized Coordinated Attention (CA) module, enhancing the lightweight network’s capability to learn feature representations. Experimental results on spaceborne SAR images demonstrate that compared to other deep generative networks focused on SAR image super-resolution reconstruction, LSRGAN achieves compression ratios of $74.68\%$ in model storage requirements and $55.93\%$ in computational resource demands. In this work, we significantly reduce the model complexity, improve the quality of spaceborne SAR images, and validate the effectiveness of the SAR image super-resolution algorithm as well as the feasibility of real-time on-board processing technology.

The improvement in azimuth resolution and/or the increase in the instantaneous beam width can result in a wide-beam autofocus problem in airborne synthetic aperture radar (SAR). Because of the assumption of central-beam approximation, the typical two-step motion compensation (MoCo) method and its extensions are unsuitable for the autofocus of wide-beam SAR imaging. In addition, existing MoCo algorithms for airborne wide-beam SAR imaging require high-precision inertial navigation data. To address these problems, we propose a wide-beam autofocus algorithm based on a blind nonlinear chirp scaling. First, based on the two-step MoCo method, the wide-beam autofocus in this study was modeled as a multiparameter optimization problem for an optimal overall imagery quality. Subsequently, we combine the well-known parametric and nonparametric methods in the SAR autofocus community to convert the multidimensional optimization problem into a two-dimensional optimization problem. Furthermore, we observe that this two-dimensional optimization problem can be further converted into two one-dimensional optimization problems, which can significantly increase the efficiency and robustness of autofocus processing. In addition, the proposed method is a full-aperture algorithm for wide-beam SAR autofocus, which can avoid the problems of image stitching and phase discontinuity compared with well-known subaperture autofocus algorithms (e.g., SATA, PTA, and FD). The processing results of the simulated and measured data verified the effectiveness of the proposed method.

For millimeter-wave radar application, a very compact planar W-band Tx/Rx front-end module is proposed in this paper. Its novel planar and integrated design benefits from a electromagnetic bandgap (EBG) packaging and the low temperature co-fired ceramic (LTCC) circuits. The unit cell of the substrate integrated EBG is proposed instead of the traditional machined nail unit cell, and its periodic structure is applied for the electromagnetic compatibility of this W-band front-end. Moreover, there are three LTCC circuits applied in this W-band front-end module, including the microstrip-to-waveguide transition, bandpass filter, and forward-wave directional coupler. Finally, the EBG packaging and LTCC circuits are fabricated. Then, the proposed planar and integrated W-band front-end module is assembled and measured with other active chips. The transmitted power is around 28.5 dBm, and the receive noise figure is around 7.5 dB, which demonstrate the feasibility of such a W-band Tx/Rx front-end module.