Yi Liang

Robust registration of multimodal remote sensing images (MRSIs) remains a challenge due to nonlinear intensity differences (NIDs) and geometric distortions. Existing methods typically rely on maximum index maps (MIMs) to construct feature descriptors. However, MIM is sensitive to rotation variations, under which circumstance the number of correct matches decreases under severe geometric distortions. To address these problems, this study proposes an ingenious registration method (named as CSFD-AFM) based on a cyclic shift feature descriptor (CSFD) and an adaptive feature matching (AFM). The CSFD is a rotation-invariant descriptor that enables efficient orientation adjustment through cyclic shifts of a three-dimensional feature array constructed from a modified MIM. To achieve rotation invariance, a dominant-orientation-based method for cyclic shift is introduced. Based on the CSFD, we devise a matching strategy named AFM, which adaptively adjusts the scale and orientation of CSFDs for robust and accurate feature matching. Experiments were conducted on six MRSI datasets to evaluate the performance of the proposed CSFD-AFM. The results suggest that the proposed CSFD-AFM outperforms the existing seven state-of-the-art methods including ReDFeat, MINIMA-LG, RIFT2, MS-HLMO, HOWP, POS-GIFT, and GLS-MIFT. In addition, our proposed CSFD-AFM achieves the accuracy within 2 pixels while maintaining its robustness against severe geometric distortions such as large rotations.

Sparse reconstruction-based imaging techniques can be utilized to solve forward-looking imaging problems with limited azimuth resolution. However, these methods perform well only under the traditional model for the platform with low speed, and the performance deteriorates for the maneuvering trajectory. In this paper, a structured Bayesian super-resolution forward-looking imaging algorithm for maneuvering platforms under an enhanced sparsity model is proposed. An enhanced sparsity model for maneuvering platforms is established to address the reconstruction problem, and a hierarchical Student-t (ST) prior is designed to model the distribution characteristics of the sparse imaging scene. To further leverage prior information about structural characteristics of the scatterings, coupled patterns among neighboring pixels are incorporated to construct a structured sparse prior. Finally, forward-looking imaging parameters are estimated using the expectation/maximization-based variational Bayesian inference. Numerical simulations validate the effectiveness of the proposed algorithm and the superiority over conventional methods based on pixel sparse assumptions in forward-looking scenes for maneuvering platforms.

In the process of airborne radar imaging, when encountering certain scenarios (e.g., the Gobi desert), the low number or absence of strong scattering points will lead to azimuth scattering in the imaging results. In this study, we propose an autofocusing method based on the traditional phase gradient autofocus (PGA) algorithm that determines the optimal number of strong scattering points in each azimuth sub-aperture according to the scene (i.e., adaptively) and uses the mean amplitudes of these points in each azimuth sub-aperture of the scene to make a threshold judgment. In this way, the optimal number of strong scattering points is iteratively selected for subsequent PGA processing. The focusing quality of the image is significantly improved through the combination of these algorithms with onboard radar imaging. The simulation results prove the effectiveness of the method proposed in this study, compared with traditional PGA.

For multifunctional phased array radar systems, inverse synthetic aperture radar (ISAR) imaging typically requires multi-dwell coherent processing to achieve adequate cross-range resolution. In such cases, the traditional fast Fourier Transform (FFT)-based Doppler processing method can result in grating lobes in the cross-range dimension. Additionally, the range resolution is insufficient when the radar transmission bandwidth is relatively narrow. To address these two issues, this paper proposes a fast sparse Bayesian learning (SBL)/iterative Wiener filter (IWF)-based two-dimensional (2D) joint super-resolution ISAR imaging method, which enhances the resolution in both range and cross-range dimensions while suppressing grating lobes. The proposed fast SBL and IWF algorithms incorporate the lower-tri-Toeplitz-cyclic (LC) decomposition of the unfolded fourth-order Toeplitz tensor of the covariance matrix. Except for the LC decomposition, all operations utilize the FFT, allowing for efficient, precise, and memory-efficient processing of two types of multi-dwell observations. Furthermore, during image reconstruction, the target’s rotational velocity is estimated using the minimum entropy criterion, thereby achieving range-variant autofocus and 2D super-resolution ISAR imaging. Finally, simulated and measured data are employed to validate the effectiveness of the proposed algorithms.

The satellite–ground bistatic configuration, which uses geosynchronous synthetic aperture radar (GEO SAR) for illumination and ground equipment for reception, can achieve wide coverage, high revisit, and continuous illumination of interest areas. Based on the analysis of the signal characteristics of GEO satellite–ground bistatic SAR (GEO SG-BiSAR), it is found that the bistatic echo signal has problems of azimuth spectrum aliasing and 2-D spatial variability. Therefore, to overcome those problems, a novel SAR imaging method for a GEO SG-BiSAR system with severe azimuth spectrum aliasing and 2-D spatial variability is proposed. Firstly, based on the geometric configuration of the GEO SG-BiSAR system, the time-domain and frequency-domain expressions of the signal are derived in detail. Secondly, in order to avoid the increasing cost caused by traditional multi-channel reception technology and the processing burden caused by inter-channel errors, the azimuth deramping is executed to solve the azimuth spectrum aliasing of the signal under the special geometric structure of GEO SG-BiSAR. Thirdly, based on the investigation of azimuth and range spatial variability characteristics of GEO SG-BiSAR in the Range Doppler (RD) domain, the azimuth spatial variability correction strategy is proposed. The signal corrected by the correction strategy has the same migration characteristics as monostatic radar. Therefore, the traditional chirp scaling function (CSF) is also modified to solve the range spatial variability of the signal. Finally, the two-dimensional spectrum of GEO SG-BiSAR with modified chirp scaling processing is derived, followed by the SPECAN operation to obtain the focused SAR image. Furthermore, the completed flowchart is also given to display the main composed parts for GEO SG-BiSAR imaging. Both azimuth spectrum aliasing and 2-D spatial variability are taken into account in the imaging method. The simulated data and the real data obtained by the Beidou navigation satellite are used to verify the effectiveness of the proposed method.

In recent years, complex-valued convolutional neural networks (CNNs) have emerged as a promising approach for polarimetric synthetic aperture radar (PolSAR) image segmentation by utilizing both amplitude and phase information in PolSAR data. This article introduces a complex-valued network for PolSAR image segmentation termed as complex-valued Lovász-softmax loss optimization synthetic aperture radar network (CV-LoSARNet), which is in fact a complex-valued Lovász-softmax loss optimization framework. The bilateral structure of CV-LoSARNet provides efficient feature extraction, while the complex-valued network adapting to PolSAR data can improve feature extraction capabilities. The introduced loss function combines both the Lovász-softmax loss and cross-entropy loss, which can improve the optimization objective of the segmentation. Comparative experiments conducted on E-SAR data and AIRSAR data demonstrate the superiority of the proposed network over the classical full CNN and the classic bilateral networks. Compared with the classic bilateral network, the CV-LoSARNet has improved the mean intersection over union and mean pixel accuracy of E-SAR data sets by 2.37% and 2.29%, for AIRSAR data sets, the improvement is 12.95% and 6.70%. Moreover, the segmentation performance of the proposed network on different polarimetric modes is discussed.

Bistatic synthetic aperture radar (BiSAR) can perform remote imaging at any position, thus playing a crucial role in remote sensing applications. In BiSAR imaging with complex trajectory (CT-BiSAR), an inevitable issue is the complicated correspondence between the geometry space and signal space, which will cause Doppler parameters mismatch and affect the imaging performance. To address this problem, the spatial inverse mapping relationship is skillfully constructed through the method of series reverse (MSR), which accurately describes the specific correspondence between the geometry space and signal space. On this basis, an extended time scaling (ETS) algorithm is proposed. Its innovation is compensating for spatially variant Doppler phase that causes image distortion and defocusing by introducing the preprocessing and scaling functions in the time and frequency domains, respectively. Simulation experiments validate the effectiveness of the proposed method.

Polar format algorithm (PFA) is a widely used high-resolution SAR imaging algorithm that can be implemented in advanced widefield staring synthetic aperture radar (WFS-SAR). However, existing algorithms have limited analysis in wavefront curvature error (WCE) and are challenging to apply to WFS-SAR with high-resolution and large-swath scenes. This paper proposes an extended polar format algorithm for joint envelope and phase error correction in WFS-SAR imaging with maneuvering trajectory. The impact of the WCE and residual acceleration error (RAE) are analyzed in detail by deriving the specific wavenumber domain signal based on the mapping relationship between the geometry space and wavenumber space. Subsequently, this paper improves the traditional WCE compensation function and introduces a new range cell migration (RCM) recalibration function for joint envelope and phase error correction. The 2D precisely focused SAR image is acquired based on the spatially variant inverse filtering in the final. Simulation experiments validate the effectiveness of the proposed method.

In this paper, the problem of inverse synthetic aperture radar (ISAR) high resolution imaging and cross-range scaling with periodically gapped coherent processing interval (CPI) caused by the radar performing multiple tasks time divisionally is addressed. In the case of periodically gapped CPI, the traditional fast Fourier transform (FFT) based ISAR imaging method is no longer applicable, and sparse reconstruction is an effective approach to solve this problem. Sparse Bayesian learning (SBL) is the most robust and accurate reconstruction method among all sparse reconstruction algorithms, and it is also an implementation of optimal estimation of random signal model under the minimum mean square error (MMSE) criterion. Based on the fact that the noise variance of radar observation can be estimated with secondary data, and the amplitudes of the target scatterers can be regarded as unknown deterministic variables, we propose another implementation method for the MMSE estimator, i.e., iterative Wiener filter (IWF), for ISAR imaging and cross-range scaling of periodically gapped data. Compared with SBL, the proposed method has the advantages of lower computational complexity, faster convergence, and higher reconstruction accuracy. Further, we design two fast IWF algorithms based on the triangular-circulant low displacement rank decomposition by utilizing Toeplitz-block-Toeplitz (TBT) structure, in which almost all operations except matrix decomposition can be calculated using FFT, which can efficiently and accurately process two different types of periodically gapped echo data. Finally, the performance of the algorithm was verified with both simulation and measured data.

The deep learning technique has already been successfully applied in the field of microwave remote sensing. Especially, convolutional neural networks have demonstrated remarkable effectiveness in synthetic aperture radar (SAR) image semantic segmentation. In this paper, a Lovász-softmax loss optimization SAR net (LoSARNet) is proposed which optimizes the semantic segmentation metric intersection over union (IOU) instead of using the traditional cross-entropy loss. Meanwhile, making use of the advantages of the dual-path structure, the network extracts feature through the spatial path (SP) and the context path (CP) to achieve a balance between efficiency and accuracy. Aiming at a polarimetric SAR (PolSAR) image, the proposed network is conducted on the PolSAR datasets for terrain segmentation. Compared to the typical dual-path network, which is the bilateral segmentation network (BiSeNet), the proposed LoSARNet can obtain better mean intersection over union (MIOU). And the proposed network also shows the highest evaluation index and the best performance when compared with several typical networks.