In the scene of remote sensing image (RSI) recognition, it is difficult to obtain a sufficient number of samples for training all categories at once. A more realistic situation is that the recognition task occurs in an open environment, with categories gradually increasing. Additionally, due to the difficulty of collecting certain data, there are only a few samples for each new category. This leads to the problem of few-shot class-incremental learning (FSCIL), where the model learns incrementally and the number of samples for incremental classes is very small, generally only a few, while the number of samples for base classes is relatively large. To address this, this paper proposes a model framework for FSCIL of RSIs, called HDCPAA. The model is mainly divided into three parts. The first part is the feature extraction network, which is pre-trained on the base classes and then its parameters are frozen in subsequent incremental learning to alleviate catastrophic forgetting of the base classes. The second part is a fully connected layer, which transforms the prototypes of each category into quasi-orthogonal prototypes to increase the distance between the prototypes. The third part is the prototype adaptation attention module, which adaptively updates prototypes and query vectors using attention mechanisms. The training process of this module is based on the meta-learning of pseudo-incremental classes. Experiments on two popular benchmark RSI datasets, MSTAR and NWPU-RESISC45, show that our model significantly outperforms the baseline models and sets new state-of-the-art results with remarkable advantages. Our code will be uploaded at: https://github.com/lipeng144/HDCPAA .

As an effective clutter suppression method for airborne radar, the atomic norm minimization (ANM)-based space-time adaptive processing (STAP) method suffers from high computational complexity and parameter setting difficulty. To solve these problems, a deep unfolded (DU) ANM algorithm is proposed for STAP in this study. First, the clutter estimation problem based on ANM is established. Then, the problem is solved via the alternating direction method of multipliers (ADMMs) and a deep neural network (DNN), which is trained by designing an appropriate loss function and constructing a complete dataset. At last, the clutter-plus-noise covariance matrix (CNCM) and the STAP weighting vector are obtained by processing the training range cell data via the trained network. Simulation results show that the proposed DU-ANM-STAP method can achieve higher clutter and noise suppression performance with lower computational cost than the existing ANM-STAP methods.

In the passive bistatic radar (PBR) system, there exists methods used to address the issue of detecting weak targets without being influenced by non-ideal factors from adjacent strong targets. These methods utilize the sparsity in the delayDoppler domain of the cross ambiguity function (CAF) to detect weak targets. However, the modeling and solving of this method involve substantial memory consumption and computational complexity. To address these challenges, this paper establishes a target detection model for PBR based on batch processing of sparse representation and recovery. This model results in a reduction of the computational complexity and memory resource requirements for sparse representation and recovery, and provides favorable conditions for parallel execution of the algorithm. Experimental results using digital video broadcasting-terrestrial (DVB-T) signal indicate that the proposed approach enables fast and stable detection of weak targets.

For airborne radar space-time adaptive processing (STAP), conventional statistic-based non-homogeneous detectors (NHDs) can help to eliminate the range cells contaminated by outliers, while their performance usually depends on the available training range cells. To solve this problem, the non-homogeneous detection problem is transformed to an out-of-distribution (OOD) data detection problem and a deep learning (DL) based NHD is then proposed in this study. According to the reconstruction errors of their corresponding radar signals in a trained complex-valued memory-augmented autoencoder, the DL-NHD can detect and eliminate the contaminated range cells and thus improve the performance of STAP effectively. Simulation results show that, compared with three existing NHDs, the proposed DL-NHD can detect outlier-contaminated range cells with 0 miss-detection and 0 false-detection. After using DL-NHD to eliminate the outlier-contaminated range cells, the STAP output signal-to-clutter-plus-noise-ratio (SCNR) loss can be improved by about 10~20 dB.

This letter introduces a novel coalitional game-theoretic power allocation (CGTPA) paradigm, tailored for resource management for communication and netted radar systems. Taking the power allocation for enhancing downlink throughput in a distributed antenna system (DAS) as a clear-cut example, theoretical expositions, and experimental simulations are presented accordingly. As an optimal decision-making method grounded in behavioral rules, the CGTPA method distinguishes itself from the methods yielded by conventional convex optimization and heuristic methods. It evolves around the idea that cooperation among antennas can be modeled as a coalitional game. Then the Sharpley value-based power allocation ensures a Pareto solution for both optimality and fairness. Furthermore, it is mathematically proven that the throughput consistently improves through iterative application of the allocation rule. Mathematical analysis and simulation results validate the effectiveness of the proposed method.

BACKGROUND: Accurate prediction of molecular property holds significance in contemporary drug discovery and medical research. Recent advances in AI-driven molecular property prediction have shown promising results. Due to the costly annotation of in vitro and in vivo experiments, transfer learning paradigm has been gaining momentum in extracting general self-supervised information to facilitate neural network learning. However, prior pretraining strategies have overlooked the necessity of explicitly incorporating domain knowledge, especially the molecular fragments, into model design, resulting in the under-exploration of the molecular semantic space. RESULTS: We propose an effective model with FRagment-based dual-channEL pretraining (FREL). Equipped with molecular fragments, FREL comprehensively employs masked autoencoder and contrastive learning to learn intra- and inter-molecule agreement, respectively. We further conduct extensive experiments on ten public datasets to demonstrate its superiority over state-of-the-art models. Further investigations and interpretations manifest the underlying relationship between molecular representations and molecular properties. CONCLUSIONS: Our proposed model FREL achieves state-of-the-art performance on the benchmark datasets, emphasizing the importance of incorporating molecular fragments into model design. The expressiveness of learned molecular representations is also investigated by visualization and correlation analysis. Case studies indicate that the learned molecular representations better capture the drug property variation and fragment semantics.

Although deep neural networks (DNNs) have achieved good results on some common datasets of synthetic aperture radar (SAR) automatic target recognition (ATR), DNNs are opaque and difficult to interpret, which limits its practical application. In view of this, many interpretable models have been proposed in recent years. However, the previous interpretable models only represent transparent network structures and cannot be called real “interpretable” models. We think that the interpretability of a model contains two meanings: on the one hand, the decision process of the model is transparent, and on the other hand, the decision-making basis of the model should be reasonable and conform to human cognition. For this reason, we propose a new SAR image recognition model based on adversarial defense, namely SAR-AD-BagNet. Not only does it have a transparent decision-making process, but it also has a more reasonable basis for decision-making. In addition, the model also has high SAR image recognition accuracy and strong adversarial robustness.

To improve target detection performance in non-Gaussian backgrounds, the joint design of transmit sequence and receive filter for multiple-input-multiple-output (MIMO) radar is studied. By approximating the probability density function of observed non-Gaussian data with the Gaussian mixture model, a Riemannian manifold of Gaussian mixture distribution is developed to depict the complicated background first. Then, maximizing the geometric distance on manifolds, which is converted by maximizing the discrimination between the target and clutter, is proposed as the criterion for the joint design of transmit sequence and receive filter. Thereby, under the constant-modulus constraint, the joint design problem can be transformed into an optimization problem. However, the proposed optimization problem is non-convex and constrained. To solve this problem, a Riemannian optimization framework is provided. By taking the advantage of the underlying geometric and algebraic structure of the constraint space, the original constrained optimization problem in Euclidean space can be transformed into the unconstraint optimization problem over Riemannian product manifolds. Moreover, to obtain the global optimal solution, the Riemannian gradient of the geometric distance cost is derived for the conjugate gradient algorithm. Experiments demonstrate that the proposed method shows advantages in detection performance compared with competitive methods.

Wideband interference (WBI) is detrimental to high-resolution radar due to its high power and wide frequency occupancy. In this study, a deep learning (DL) method is proposed to predict the time–frequency (TF) feature of WBI and applied to cognitive radar high-resolution range profile (HRRP) estimation. Specifically, by performing short-time Fourier transform (STFT) on the WBI signal collected in the past and using a sliding window, a series of WBI TF figures is generated. A long short-time memory (LSTM) network is then used to learn the spatiotemporal (ST) correlation of these TF figures, thus predicting the WBI TF feature in the future, based on which, a cognitive method is used for target HRRP estimation with reduced influences of WBI. Numerical results demonstrate the effectiveness of the proposed methods.

Microwave imaging with large rotating angle and sparse sampling is an attractive approach to obtain the high-resolution target image with reduced radar resource. However, the popular imaging methods, e.g., Range-Doppler (RD), back projection (BP), and sparse recovery (SR), are difficult to deal with large rotating angle and sparse sampling simultaneously. In recent years, deep learning (DL) has been widely studied and been successfully used to handle the problems in computer vision. However, since most existing DL networks are put forward for the real visual image and a large amount of data is essential for network training, DL cannot be directly used to process the complex and sparse target echo for microwave imaging. In this article, a new learning imaging framework is proposed and a model-driven learning imaging network (MDLI-Net) is built for high-resolution microwave imaging with large rotating angle and sparse sampling. In the proposed framework, the electromagnetic scattering model is used to generate the training data efficiently, and the sparse microwave imaging theory is applied to guide the design of the deep imaging network. By inputting the 2-D sparse complex-valued target echo, the trained MDLI-Net can output the high-resolution and focused target image efficiently. The effectiveness of the proposed learning imaging method is validated by experiment results with both simulated and real data.

A monostatic/bistatic ground-based synthetic aperture radar (GB-SAR) system using an optical electric field sensor (OEFS) as the bistatic receiving unit has been designed for target imaging and 2-D displacement estimation purposes. The designed system has the capability of acquiring the monostatic and bistatic SAR images simultaneously from different looking angles, enabling the displacement vector estimation of the illuminated area. The spatial baseline between the monostatic and bistatic receivers is selected according to the estimation precision analysis. Experimental results are presented to evaluate the performance of the designed system and the proposed method. The displacement estimation accuracies in x- and y-directions can reach up to millimeter level.

Stepped frequency continuous wave (SFCW)-based radar with passive cooperative target is a promising nondestructive method to monitor the concrete temperature change. By burying a surface acoustic wave (SAW) sensor in the concrete to act as the cooperative target, the physical properties of the concrete can be measured by analyzing the reflections of the probing SFCW signal. Because the SAW sensor should be connected to an antenna to convert the electromagnetic wave into the acoustic wave, a half-cut compact monopole antenna is designed and fabricated in this letter. Taking the advantages of corrugated edge and half-structure technologies, the size of the monopole antenna decreases significantly without sacrificing the performance, which makes the SAW sensor together with antenna suitable to be inserted into the concrete. The experimental results show that the proposed antenna can work well with the SAW sensor in the concrete and the proposed method can measure the temperature change in concrete continuously.

In this study, we used the digital television signal broadcasted by geostationary satellite for passive bistatic synthetic aperture radar (SAR) imaging and interferometry applications. A prototype system was designed and fabricated with commercial-off-the-shelf (COTS) components. Specifically, three COTS low noise block downconverters (LNBs) were synchronized for a longtime coherent measurement. Multiple TV channels were combined to achieve a wider bandwidth to improve the range resolution. The potential applications of this ground-based radar system are SAR imaging, displacement estimation, and digital elevation model generation. Hardware design and data processing algorithms are described. Experimental results are presented to validate the performance of the designed system and the proposed methods.

In this study, a ground-based radar system with two multiple input multiple output (MIMO) arrays is developed for 2D displacement estimation applications. To reduce the influences of antenna position error, channel delay, and phase/gain difference on the imaging quality, calibration methods and phase coherence factor (PCF) based filtering methods are applied. For each MIMO array, target displacement can be measured along its line of sight direction. Co-registration is conducted to help to derive the 2D displacement vector estimations of the targets.

A novel ground-based bistatic polarimetric synthetic aperture radar (SAR) system using an Optical Electric Field Sensor (OEFS) as the receiver was developed for environmental studies. Fundamental experiments were carried out with a trihedral corner reflector (CR) as a target, showing the polarimetric capability of the system. Different polarization signatures from the monostatic SAR images were found by analyzing the bistatic polarimetric SAR images. Meanwhile, this system has the capability to estimate the two-dimensional (2D) displacement of the targets in the field of view. The simulation shows the feasibility and effectiveness of the 2D displacement estimation. The experimental results demonstrate that the accuracy along the x and y direction to the line of sight (LOS) based on the designed bistatic synthetic aperture radar system can reach millimeter level for the displaceable corner reflector.

We designed a ground-based radar system with a C-band 2-D cross multiple input multiple output (MIMO) array for 3-D imaging and displacement estimation purposes. For this system, we developed a far-field pseudo-polar image format algorithm using pseudo-polar spherical coordinate. The use of a tensor compressive sensing technique allows to focus under-sampled raw data and to optimize the data acquisition time and memory usage. A novel algorithm, named as tensor-based iterative adaptive approach, is proposed for the effective and efficient reconstruction of sparse targets with a reduced level of sidelobes. Experimental results validate the designed radar system and the proposed algorithms.

To reduce the data acquisition time and the high-level sidelobes produced by conventional focusing methods for ground-based synthetic aperture radar interferometry, we present a new method to provide accurate displacement maps based on the dimension-reduced compressive sensing (CS) method combined with the multiple measurement vectors (MMVs) model. The proposed CS method consists in selecting the supported area of targets, estimated by the fast conventional method with undersampled data. The following sparse reconstruction is applied only to the selected areas. The MMV-based approach allows increasing the coherence and the precision of displacement estimates. Two experiments are carried out to assess the performance of the proposed method.

We have studied passive radar with digital broadcasting terrestrial signals for imaging moving and stationary targets. We combined multiple TV channels to improve the range resolution. In order to reduce the high-level sidelobes caused by the frequency gaps among multiple channels, a low rank matrix completion based method is proposed. We present the experiment results of range- Doppler mapping of moving targets, high-resolution time delay estimation, and passive synthetic aperture radar (SAR) imaging of stationary targets. It is shown that, by the designed system, a landing airplane can be detected and tracked. By combining multiple TV channels, high resolution time delay estimation can be achieved. Buildings located within 55 meters can be effectively imaged by the proposed method.

We study the imaging and interferometric performances of a MIMO radar on board of an airship as alternative to airborne and spaceborne SAR remote sensing techniques. Four different MIMO radar arrays are designed working in L, C, X and Ku frequency bands. A frequency bandwidth of 250 MHz has been considered for the MIMO radars. The spatial resolution is 0.6 m in range and 0.3 degree in azimuth. The imaging and interferometric performances of the MIMO radar are analyzed in terms of the airship stability. A synthetic raw data set is generated assuming a target deployed on a flat area at different azimuth angle. This MIMO imaging solution is intended for continuous imaging over an area of interest.

In actual condition, array elements deficiency or transmission errors lead to incomplete data, which is called sparse aperture (SA) data. In inverse synthetic aperture radar (ISAR) imaging, this large‐gaped data produces poor‐quality ISAR images when using traditional range–Doppler algorithm. Recently, imaging algorithms based on compressed sensing (CS) theory alleviate this problem effectively because CS theory indicates that sparse signal can be reconstructed from incomplete measurements. However, the basis mismatch problem in CS‐based algorithms may degrade the ISAR image. In this study, a reweighted atomic‐norm minimisation (ANM) (RAM)‐based imaging method is proposed. RAM is a gridless sparse method, which can enhance sparsity and resolution. RAM formulates an optimisation problem and iteratively carries out ANM with a sound reweighting strategy. By reformulating the RAM as a semi‐definite programme, the echoes with full aperture (FA) are reconstructed from SA data. After that, ISAR imaging with the reconstructed FA data is achieved via the conventional azimuth compression method. Simulated and real data results demonstrate the effectiveness and superiority of the proposed method.

A novel compressive sensing (CS) based imaging method for SFCW radar near range targets is proposed. Different from existing CS based methods, the azimuth-dependency of the target reflection coefficient is considered. Based on the block sparsity property of the received signal in the proposed sparsifying dictionary, the 2D image of targets can be obtained at each spatial sampling point. Cross-correlation method is then employed to fuse these 2D images to get the final result. Compared to the classical back projection (BP) method, the proposed method can obtain higher resolution with fewer artifacts via fewer frequencies. Compared to the conventional CS based method, artifacts can be significantly reduced. Experiment results of a SFCW linear sparse array radar system demonstrate that, with only 1/8 data, the proposed method can achieve accurate high-resolution 2D image of targets in the near range. In 30dB dynamic range, no artifact was produced on the imaging result.