Low Earth orbit (LEO) satellite is one of the most promising infrastructures for realizing next-generation global wireless networks with enhanced data rates. Applying massive multiple-input multiple-output (mMIMO) to LEO satellite communication systems is a novel idea to enhance communication capacity and realize the global high-speed interconnection. However, obtaining effective instantaneous channel state information (iCSI) is challenging due to the time-varying propagation environment and long transmission delay. In this letter, a deep learning (DL)-based CSI prediction scheme is proposed to address channel aging problem by exploiting the correlation of changing channels. Specifically, we design a satellite channel predictor (SCP) that is composed by long short term with memory (LSTM) units. The predictor is first trained by offline learning and then feeds back the corresponding output results online based on the input data to realize channel feature extraction and future CSI prediction in LEO satellite scenarios. Numerical results demonstrate that the proposed DL-based predictor can mitigate the channel aging problem in LEO satellite mMIMO system effectively.

In an azimuth multichannel synthetic aperture radar (SAR) system, the unavoidable channel error will result in virtual targets or azimuth ambiguity, which significantly degrades the quality of high-resolution and wide-swath (HRWS) SAR image. Calibration of the imbalances between channels has been an important topic. First, to present the channel error effect on the final image clearly, a precise relation between the amplitudes of virtual targets and amplitude–phase error is established by matrix trace, which applies to the case of a certain amplitude–phase error. In addition, the total amplitude of all targets represented by the ${L^{1}}$ -norm reaches the minimum when there is no channel error. Based on this principle, an image-domain least ${L^{1}}$ -norm method is proposed to estimate the phase error between channels. By utilizing the focused high signal-to-noise ratio (SNR) images, the proposed algorithm achieves high estimation accuracy. Moreover, no redundant channel is required in the proposed algorithm compared with the subspace-based method. Finally, the effectiveness of the proposed channel error effect analysis and calibration method is validated by both the simulated and real SAR data.

Accurate and efficient image mosaicking is essential for generating wide-range swath images of spaceborne scanning synthetic aperture radar (ScanSAR). However, the existing methods cannot guarantee the accuracy and efficiency of stitching simultaneously, especially when mosaicking multiple large-area images. In this paper, we propose a novel image mosaic method based on homography matrix compensation to solve the mentioned problem. A set of spaceborne ScanSAR images from the Gaofen-3 (GF-3) satellite were selected to test the performance of the new method. First, images are preprocessed by an improved Wallis filter to eliminate intensity inconsistencies. Then, to reduce the enormous computational redundancy of registration, the overlapping areas of adjacent images are coarsely extracted using geolocation technologies. Furthermore, to improve the efficiency of stitching and maintain the original information and resolution of images, we deduce a compensation of homography matrix to implement downsampled images registration and original-size images projection. After stitching, the transitions at the edges of the images were smooth and seamless, the information and resolution of the original images were preserved successfully, and the efficiency of the mosaic was improved by approximately one thousand-fold. The validity, high efficiency and reliability of the method are verified.

P- and L-band synthetic aperture radar (SAR) sensors have foliage- and ground-penetrating imaging capabilities and thus contribute to missions such as deforestation monitoring, biomass estimation, and deformation measurements. The spectrums they employed are also allocated to many services, such as radio navigation satellite services, radio location services, telecommunication, and broadcasting, which is prone to generate harmful radio frequency interference (RFI) to SAR including time stationary narrow band (TSNB) and intermittently transmitted wideband (ITWB) versions. Researchers generally perform on-ground processing to mitigate RFIs by which notch filtering is the best known. However, the phenomena of side-lobe uplift and SNR degradation caused by spectral gaps will deteriorate SAR image quality and further affect image interpretation. This article models and characterizes the signature of TSNB and ITWB RFI in P- and L-band SAR, and on this basis, a two-stage approach is developed to mitigate RFI that overcomes the drawbacks of notch filtering-based approaches. In the first stage, a spectrum cancellation-based RFI detector is proposed, which simultaneously locates RFI and produces an orthogonal projection operator. In the second stage, a low-rank approximation model regulated by truncated nuclear norm, which is solved by the alternating direction method of multipliers (ADMM), is established to recover the missing spectral information. Experiments are conducted extensively on simulated data as well as real-world data collected by Advanced Land Observing Satellite (ALOS) Phased-Array L-Band Synthetic Aperture Radar (PALSAR) system and a P-band airborne SAR system to validate the proposed approach.

Phase denoising constitutes a critical component of the synthetic aperture radar interferometry (InSAR) processing chain, where noise suppression and detail preservation are two mutually constraining objectives. Recently, deep learning has attracted considerable interest due to its promising performance in the field of image denoising. In this paper, a Neighbor2Neighbor denoising network (NBDNet) is proposed, which is capable of simultaneously estimating phase and coherence in both single-look and multi-look cases. Specifically, repeat-pass PALSAR real interferograms encompassing a diverse range of coherence, fringe density, and terrain features are used as the training dataset, and the novel Neighbor2Neighbor self-supervised training framework is leveraged. The Neighbor2Neighbor framework eliminates the necessity of noise-free labels, simplifying the training process. Furthermore, rich features can be learned directly from real interferograms. In order to validate the denoising capability and generalization ability of the proposed NBDNet, simulated data, repeat-pass data from Sentinel-1 Interferometric Wide (IW) swath mode, and single-pass data from Hongtu-1 stripmap mode are used for phase denoising experiments. The results demonstrate that NBDNet performs well in terms of noise suppression, detail preservation and computation efficiency, validating its potential for high-precision and high-resolution topography reconstruction.

Multibaseline (MB) synthetic aperture radar inter-ferometry (InSAR) is capable of reconstructing steep terrain height profiles, and MB phase unwrapping (PU) is one of the most critical and challenging steps in its processing chain. Existing MB PU algorithms usually suffer from poor noise robustness or low computational efficiency. In this work, we propose a fast and robust MB PU algorithm, including three main steps: 1) construct the triangulation network guided by the quality map; 2) obtain robust estimation of ambiguity number gradients based on the intrinsic relationship between interferograms and gradient regularization; 3) solve minimum cost flow problem with modified weights. The mean absolute errors of the height profiles constructed by the proposed method are 1.6m and 1.3m for simulated data and real data experiments respectively, verifying the effectiveness of the proposed method.