Xiongpeng He

Ground Moving Target Indication (GMTI) has always been one of the key tasks in synthetic aperture radar (SAR) systems. For SAR-GMTI, existing GMTI algorithms can be classified into traditional signal processing algorithms and low-rank matrix recovery algorithms. Space-time adaptive processing (STAP), as a typical traditional approach, has shown excellent performance in detecting moving targets in real applications; low-rank matrix recovery algorithms have become one of the recent research hotspots. Robust principal component analysis (RPCA), as a typical low-rank matrix recovery approach, has been applied in SAR-GMTI due to its ability to decompose an approximate low-rank matrix into a low-rank component and a sparse component. However, a drawback of RPCA algorithms is its relatively high false alarm rate caused by the energy leakage from strong clutter. The detection performance of the STAP algorithm may degrade when the training samples are contaminated. To address these challenges, we propose a block-based RPCA-STAP algorithm that integrates traditional adaptive clutter suppression methods into the RPCA framework. The proposed algorithm first implements clutter classification using the Markov random field (MRF) image segmentation algorithm. Then, RPCA-STAP suppression is applied to different clutter blocks, which not only satisfies the IID requirement for STAP training samples but also reduces the false alarm rate caused by strong clutter in the RPCA algorithm by introducing STAP. Additionally, considering the sensitivity of GMTI algorithms to channel errors, we propose an improved adaptive 2-D calibration (IA2DC) algorithm to further enhance channel correlation. Simulation and real data experiments validate the effectiveness of the proposed algorithms.

In recent years, the low-rank matrix recovery theory has acquired widespread application in the radar system. For multi-channel synthetic aperture radar system, the robust principal component analysis (RPCA) has proven to be a valuable technique for effectively distinguishing moving targets from static background clutter within the image domain. However, in non-ideal environments, the RPCA is susceptible to channel errors and strong clutter, resulting in degraded target detection performance. To resolve this issue, a slow ground moving target indication (GMTI) processing algorithm is proposed in this paper. Firstly, the sample selection and data reconstruction (DR) are used to furtherly compensate channel imbalance error and registration error. Next, a RPCA optimization framework is proposed to mitigate the issue of elevated false alarm rates caused by heterogeneous environments, and the sparse matrix is obtained through the application of the alternating direction method of multipliers (ADMM). The proposed optimization model not only avoids excessive punishment of large singular values by kernel norm weighting, but also further improves the performance of target detection by introducing a difference matrix and a Fourier matrix. Finally, the estimation of the target’s radial velocity is accomplished through the utilization of the adaptive match filtering (AMF) algorithm. Compared with the traditional RPCA algorithm, the proposed algorithm significantly reduces the false alarm rate under the background of strong clutter. Theoretical analyses and measured data results verify the effectiveness of the proposed algorithm.

In dual-channel synthetic aperture radar (SAR) systems, the estimation of the four-dimensional motion parameters of the ground maneuvering target is a critical challenge. In particular, when spatial degrees of freedom are used to enhance the target’s output signal-to-clutter-plus-noise ratio (SCNR), it is possible to have multiple solutions in the parameter estimation of the target. To deal with this issue, a novel algorithm for estimating the motion parameters of ground moving targets in dual-channel SAR systems is proposed in this paper. First, the random sample consensus (RANSAC) and modified adaptive 2D calibration (MA2DC) are used to prevent the target’s phase from being distorted as a result of channel balancing. To address range migration, the RFRT algorithm is introduced to achieve arbitrary-order range migration correction for moving targets, and the generalized scaled Fourier transform (GSCFT) algorithm is applied to estimate the polynomial coefficients of the target. Subsequently, we propose using the synthetic aperture length (SAL) of the target as an independent equation to solve for the four-dimensional parameter information and introduce a windowed maximum SNR method to estimate the SAL. Finally, a closed-form solution for the four-dimensional parameters of ground maneuvering targets is derived. Simulations and real data validate the effectiveness of the proposed algorithm.

Correcting the range migration of moving targets and compensating for the coupled phase errors are crucial for ground moving target imaging (GMTIm). Most methods achieve target focusing through parameter search operations with substantial computational complexity. Additionally, the inability to address energy spreading caused by the higher-order motion futher limits the applicability of these methods. To overcome these issues, this paper proposes a novel refocusing method without motion parameter estimation for arbitrarily moving ground targets. Firstly, a range frequency difference (RFD) function without parameter estimation is constructed in the range frequency domain. Then, by conjugate multiplication with the RFD function, the coupling between range frequency and slow time of target can been removed. With an appropriate frequency interval selected, the target can also be azimuthally focused. In addition, some practical factors in applications are analyzed in detail. Compared to the traditional methods, the proposed method effectively addresses range migration and azimuth defocusing caused by higher-order phase errors. Also, it achieves target focusing without search operations, despite the existence of Doppler center ambiguity and Doppler spectrum splitting. Additionally, since it only requires fast Fourier transform (FFT), inverse fast Fourier transform (IFFT) and matrix multiplication operations, this method achieves high computational efficiency. The efficacy of the proposed method is confirmed through the examination of both simulated and real data.

This letter proposes a time-diverse wide-swath imaging radar transmit waveform optimization design method. Firstly, based on the imaging geometry and zebra maps, we obtained the angles corresponding to the range blind zone. Then, utilizing the mapping characteristics between the range frequency and beam scanning angle in time diverse array (TDA) radar, as well as the blind region information, we performed a two-dimensional optimization design of the transmit waveform in the fast time and range frequency domain. Finally, the limited energy can be effectively skipped over the blind regions and flexibly allocated to the observable areas. Compared to the traditional TDA system, this method achieves a larger imaging swath and energy utilization efficiency. The effectiveness of the proposed method is verified through simulation experiments.

Video synthetic aperture radar (SAR) has exhibited considerable potential for detecting and tracking ground moving targets. Numerous classical shadow-based detection methods have been applied in video SAR. In addition, shadow-assisted detection methods based on convolutional neural networks (CNNs) have been developed. In this article, we propose a joint detection method for moving targets in video SAR, which can combine the information of the video SAR image and the corresponding sparse image to suppress background interference sufficiently and improve detection accuracy. Specifically, the low-rank sparse decomposition technology is first applied for video SAR images to generate their corresponding sparse images in which the background is eliminated and shadows of moving targets are enhanced. Then, we improve faster RCNN and build a two-stream extraction feature network based on the Transformer structure that allows the video SAR image and the sparse image as input simultaneously as well as extracts and fuses the features from two types of the images, which can acquire more discriminative target features, thereby improving the final the detection performance. Furthermore, the improved faster RCNN only modifies the original feature extraction network. Thus, it can adopt the same training and test manner as faster RCNN, greatly facilitating its utilization. Finally, experiment results on Sandia National Laboratories data demonstrate that the proposed detection method outperforms other state-of-the-art methods. And our method reduces the false alarms by 1.02%, the missed detections by 43.24%, and increases the mean average precision by 2.98%.

Few-Shot Industrial Anomaly Detection (FS-IAD) has drawn great attention most recently since data efficiency and the ability to design algorithms for fast migration across products have become the main concerns. The difficulty of memory-based IAD in low-data regime primarily lies in inefficient measurement between the memory bank and query images. We address such a pivotal issue from a new perspective of optimal matching between features of image regions. Taking the unbalanced nature of query features into consideration, we adopt Conditional Transport (CT) as a metric to compute the structural distance between representations of the two sets to determine feature relevance. CT distance generates the optimal matching flows between unbalanced structural elements that achieve the minimum matching cost, which can be directly used for IAD since it well reflects the differences of query images compared with the normal memory. Realizing the fact that query images usually come one-by-one or batch-by-batch, we further propose an Online Conditional Transport (OCT) by making full use of the current and historical query images for IAD via simultaneously calibrating the memory bank using the online query images and matching features between the calibrated memory and the current query image. Go one step further, for sparse foreground products, we employ a predominant segment model to implement Foreground-aware OCT (FOCT) for improving the effectiveness and efficiency of OCT by forcing the model to pay more attention to diverse targets rather than redundant backgrounds when calibrating the memory bank. FOCT can improve the diversity of calibrated memory, which is critical for robust FS-IAD in practice. Besides, FOCT is flexible since it can be friendly plugged and played with any pre-trained backbones, such as WRN, and any pre-trained segment models, such as SAM. The effectiveness and efficiency of our model is demonstrated across diverse datasets, including benchmarks of MVTec and MPDD, achieving SOTA performance.

Video synthetic aperture radar (video SAR) has been successfully applied in many fields. And the registration of the video SAR images has been proven to be a crucial step in their preprocessing. However, video SAR images exhibit more severe image differences because of the unique imaging mechanism and the immature imaging methods. This results in existing registration methods failing to achieve satisfactory registration outcomes for video SAR images. The convolutional neural network (CNN) can contribute to improving registration performance. Nevertheless, CNN-based registration methods must be driven by a large amount of labeled data, which is impractical for video SAR images. Therefore, to tackle these problems, we propose an unsupervised end-to-end deep registration method for video SAR images. Firstly, an end-to-end deep registration model (DRM) is proposed to improve the registration performance for video SAR images. In the proposed DRM, the offset field is utilized to indirectly calculate the registered parameters. And we construct a CNN, MUnet, to regress the offset field accurately. We also develop a differentiable H-transform (DHT) and a differentiable spatial transformation (DST) to implement the mapping from end to end while allowing DRM to backpropagate the losses during the training phase. Meanwhile, we borrow intensity-based methods to further optimize the registration results. Furthermore, we propose an unsupervised deep training strategy that can use the generated pseudo data with pseudo label to train the proposed DRM in the absence of large amounts of labeled data. Experiment results on multiple data demonstrate the effectiveness of the proposed registration method.