In synthetic aperture radar (SAR) target recognition, the interpretation of SAR images and the accurate recognition of targets are significantly affected by speckle noise. Traditional recognition methods often fail to meet the high accuracy requirements. Therefore, this letter proposes a dual-ended fusion network (DEFNet) for SAR target recognition. The network consists of three main components: the local refinement feature extraction (LRFE) network, the large-scale feature extraction (LSFE) network, and the adaptive feature interaction fusion (AFIF) module. It employs a parallel structure, utilizing the LRFE and LSFE branch modules to extract local and large-scale features, respectively. These two features are then adaptively fused through the AFIF module to further enhance feature representation. Experimental results on the moving and stationary target acquisition and recognition dataset indicate that DEFNet demonstrates significant performance improvement compared to traditional methods, showcasing its effectiveness and adaptability in SAR target recognition tasks.

Traditional radar signal modulation recognition (RSMR) methods struggle to achieve the required accuracy under low signal-to-noise ratio (SNR) conditions. To address this issue, a hybrid network architecture integrating deformable convolution and mamba (DCMNet) is proposed. Specifically, DCMNet employs a multi-view feature extraction structure that combines inverted deformable convolution (IDC) with a state space model (SSM), enabling dynamic adjustment of convolution kernel positions and capturing global information and dependencies in long sequence data. The cross-gated feature fusion (CGFF) mechanism effectively modulates and dynamically aggregates features from different perspectives. The lightweight design provides significant advantages in terms of network scale and deployment. Experimental results demonstrate that the proposed method achieves excellent performance on a dataset with ten different waveforms. Notably, at an SNR of -8 dB, the recognition accuracy exceeds 90%, significantly outperforming existing methods.

Conventional sparse uniform arrays (SUAs) is composed of multiple identical and rigorously collinear uniform linear arrays. By adjusting the baseline length between the subarrays, the array aperture can be arbitrarily large, thus substantially improving the accuracy of the direction-of-arrival (DOA) estimation. However, in practical applications, it is challenging to meet the strict collinearity requirement due to geographical constraints. In this letter, to address this problem, we propose the non-collinear sparse uniform array (NCSUA) model to mitigate the influence of the non-ideal terrain and enhance the practicality of the SUA. A novel estimation algorithm is then proposed to resolve the angle ambiguity in NCSUA and effectively achieve high-accuracy DOA estimation. Compared with the conventional SUA, numerical simulation results demonstrate the superiority of NCSUA employing the new de-ambiguity algorithm in DOA estimation performance and practical applications.

To tackle the issue of space-time adaptive processing (STAP) performance degradation caused by inaccurate estimation of the clutter covariance matrix (CCM) with limited sample support, a knowledge-aided (KA) CCM estimation algorithm based on the symmetric alternating direction method of multiplier (KA-SADMM) is proposed. The CCM estimation problem is constructed based on the knowledge of persymmetric structure, the low-rank structure, and the prior covariance matrix, and the solution to the resulting problem is derived. Moreover, the contraction property of the sequence generated by KA-SADMM with respect to the solution set of the estimation problem is analyzed. With limited training samples, the simulation results show that the proposed algorithm improves the STAP performance over other similar algorithms by at least about 0.2 dB when the prior knowledge is accurate and by at least 0.3 dB when the prior knowledge is inaccurate.

Excellent performance in supervised learning-based radar signal modulation recognition (RSMR) techniques relies on the quantity and quality of labeled datasets, while the high cost and difficulty involved in analyzing and labeling radar signal samples limits its development. An RSMR algorithm using self-supervised contrastive learning (SSCL) methodology is proposed to address this issue. Specifically, within the classical contrastive learning (CL) framework MoCo V2, a customized data augmentation method is devised to capture time-frequency features of the radar signal. In addition, the feature extraction network ResNet50 is improved by decoupling spatial and channel filters, resulting in greater sensitivity to the time-frequency features. To enhance the recognition accuracy, two loss functions, alignment and uniformity, are used in place of the info noise contrastive estimation (InfoNCE) loss, and both loss functions are optimized directly. The recognition accuracy of the proposed method can reach 97.66% at a signal-to-noise ratio (SNR) of 4 dB.

The digital ubiquitous radar enhance the echo of small target by long-term integration, but the tracking of UAV is still affected by target motion patterns, sea clutter interference and other factors, which may resulting in Non Gaussian noise with significant variation. A joint optimization of kernel width and process noise covariance matrix is proposed in Cauchy kernel-based extend Kalman filter to solve this problem. By setting the kernel width as a function of the error, iteration of the kernel width is added to the algorithm so that the error decays the fastest along the rising gradient, and then the process noise covariance matrix is corrected to serve as the basis for the estimation of the next moment.Simulation and tracking experiment demonstrate that the proposed algorithm exhibits better performance.In complex noise environments, the RMSE of the algorithm is reduced by 14.13% compared to EKF.

The research on maritime target detection signals has significant value in various fields. Conventional statistical theory-based target detection methods are limited by the complex sea clutter environment and target characteristics, making it challenging to achieve high-performance detection. In practical scenarios such as maritime observation, the radar observation area is expansive. And the radar beam cannot remain fixed in one direction for prolonged intervals. Consequently, it is not feasible to accumulate multiple pulses within a single azimuth cell. Therefore, extraction of effective features from echo signals is not practical. To address this issue, this paper proposes a maritime target detection method based on the Spatial-Temporal Feature Attention Graph Convolutional Network (STFA-GCN) and radar signal graph data. Firstly, the multi-frame radar signal is converted into graph data to represent spatial-temporal features. Then a STFA-GCN model perform feature extraction and classification on the graph data nodes, realizing target detection in complex sea clutter backgrounds. The proposed method was tested and evaluated using various target datasets, exhibiting superior detection performance and generalization capabilities. On real measured signal test, the proposed method can achieve 0.917 detection probability at false alarm rate of 1.26×10-4. While the 3-frame accumulation CACFAR is 0.839 at 1.65×10-4.

Multiple Input Multiple Output (MIMO) radar, a new radar system with waveform diversity, can improve detection performance. However, there are still challenges in the current MIMO radar target detection process, such as difficult waveform separation, high data demand, high algorithm complexity, and poor detection performance. To address these issues, this letter presents a Space-Time-Waveform Joint Adaptive Detection (STWJAD) method. By combining spatial, temporal, and waveform dimensions, the STWJAD is based on the Linearly Constrained Minimum Variance (LCMV) criterion to achieve effective adaptive processing and detection. Experimental results demonstrate that the proposed method can effectively suppress sidelobes, clutter and noise, exhibit excellent detection capabilities, and boast a lower data demand, faster processing speed.

Radar is an extremely valuable sensing technology for detecting moving targets and measuring their range, velocity, and angular positions. When people are monitored at home, radar is more likely to be accepted by end-users, as they already use WiFi, is perceived as privacy-preserving compared to cameras, and does not require user compliance as wearable sensors do. Furthermore, it is not affected by lighting conditions nor requires artificial lights that could cause discomfort in the home environment. So, radar-based human activities classification in the context of assisted living can empower an aging society to live at home independently longer. However, challenges remain as to the formulation of the most effective algorithms for radar-based human activities classification and their validation. To promote the exploration and cross-evaluation of different algorithms, our dataset released in 2019 was used to benchmark various classification approaches. The challenge was open from February 2020 to December 2020. A total of 23 organizations worldwide, forming 12 teams from academia and industry, participated in the inaugural Radar Challenge, and submitted 188 valid entries to the challenge. This paper presents an overview and evaluation of the approaches used for all primary contributions in this inaugural challenge. The proposed algorithms are summarized, and the main parameters affecting their performances are analyzed.

As a classic deep learning target detection algorithm, Faster R-CNN (region convolutional neural network) has been widely used in high-resolution synthetic aperture radar (SAR) and inverse SAR (ISAR) image detection. However, for most common low-resolution radar plane position indicator (PPI) images, it is difficult to achieve good performance. In this paper, taking navigation radar PPI images as an example, a marine target detection method based on the Marine-Faster R-CNN algorithm is proposed in the case of complex background (e.g., sea clutter) and target characteristics. The method performs feature extraction and target recognition on PPI images generated by radar echoes with the convolutional neural network (CNN). First, to improve the accuracy of detecting marine targets and reduce the false alarm rate, Faster R-CNN was optimized as the Marine-Faster R-CNN in five respects: new backbone network, anchor size, dense target detection, data sample balance, and scale normalization. Then, JRC (Japan Radio Co., Ltd.) navigation radar was used to collect echo data under different conditions to build a marine target dataset. Finally, comparisons with the classic Faster R-CNN method and the constant false alarm rate (CFAR) algorithm proved that the proposed method is more accurate and robust, has stronger generalization ability, and can be applied to the detection of marine targets for navigation radar. Its performance was tested with datasets from different observation conditions (sea states, radar parameters, and different targets).

Due to the complex sea clutter environment and target features, the conventional statistical theory-based methods cannot achieve high performance in maritime target detection tasks. Conventional deep learning, such as convolutional neural networks (CNNs)-based target detection methods process each signal sample independently, and the temporal-spatial domain correlation information is seldom used. To achieve full utilization of information contained in radar signals and improve the detection performance, a graph convolutional network (GCN) is considered, which has shown great advantages in graph data processing and has been applied in the field of signal processing. This letter proposed a maritime target detection method based on radar signal graph data and graph convolution. Graph structure data is applied to define the detection units and to represent the temporal and spatial information of detection units. The target detection of the signal corresponding to the nodes is conducted via GCN. Experimental results show that the proposed approach can effectively detect marine targets when the signal-to-noise ratio is above −5 dB and can effectively suppress false alarms in the pure clutter area, which is not adjacent to targets. Compared with the popular used CNN method, e.g., LeNet, the proposed method can achieve higher detection probability with the same given false alarm rate.

Small target detection in high-resolution sea clutter is one of the important problems in radar target detection. We propose a new detector that adopts the visibility graph (VG) algorithm and the graph convolution neural network (GCN). This detector extracts the graph feature of radar echo networks for target detection after converting the phase information of radar echo into complex networks. Different from the existing detector based on graph connectivity, the VG algorithm does not need radar data preprocessing and it can directly convert the radar data into complex networks by judging whether the series data are visible mutually. We extracted the graph feature that can effectively distinguish the background clutter and target echo by GCN instead of calculating the statistic of graphs manually. Experiments on the IPIX radar datasets show that the proposed method achieves superior performance than the existing detector based on graph connectivity, especially in a short observation time. When the observation time is 0.256 s, the average accuracy of #17, #54, and #320 datasets can reach 84.82%, 94.84%, and 90.59%, respectively.

The conventional pulse compression method [i.e., matched filter (MF)] and adaptive pulse compression (APC) methods are based on the assumption that the echo sampling point is located in the target point which is in a range cell. If the echo sampling point is not located in the target point, it will result in sampling mismatch problem and the performance of MF and APC methods will be significantly reduced when using the continuous-time phase-modulated waveforms, such as linear frequency modulation (LFM) signal. Aiming at solving the sampling mismatch problem, an APC based on two-stage phase compensation (TPC-APC) and its dimensionality-reduced version-contiguous fast APC based on two-stage phase compensation (TPC-CFAPC) are proposed in this article. In TPC-APC/TPC-CFAPC method, the mismatch phase caused by the sampling mismatch is first compensated through the first-stage phase compensation, which suppresses the sampling-mismatch-induced range sidelobes; then the mismatch phase caused by the target’s Doppler frequency is compensated through the second-stage phase compensation to suppress the Doppler-mismatch-induced range sidelobes; finally, the APC is applied to suppress the range sidelobe. To further compensate the mismatch phases of different targets, this article proposes a reiterative TPC-APC method and its dimensionality-reduced version, which iteratively use TPC-APC or TPC-CFAPC to suppress the range sidelobes caused by the mismatch phases of different targets. The results of the experiments have shown that the proposed methods are more robust compared with MF and APC methods.

The problem of strong sea clutter, e.g., sea spikes, may bring in low signal-to-clutter ratio (SCR) and cause great interference to radar marine target detection. However, the sea clutter suppression ability of current algorithms is limited with poor generalization under complex marine environment. In this letter, a novel sea clutter suppression generative adversarial network (SCS-GAN) is designed and employed for marine radar plan-position indicator (PPI) images detection. The SCS-GAN is based on residual networks and attention module, which includes residual attention generator (RAG) and sea clutter discriminator (SCD). In order to expand the data sets and improve generalization ability, clutter-free data set A, simulated sea clutter data set B (containing five types of sea clutter distributions), and actual sea clutter data set C are constructed by means of simulation and acquisition of real radar returns. At last, the parameter, i.e., clutter suppression ratio (CSR) is designed for evaluating the sea clutter suppression performances of the proposed method and other denoising and clutter suppression methods including CBM3D, denoising convolutional neural network (DnCNN), FFDNet, and Pix2pix. After testing with actual data, it is proved that the SCS-GAN has faster clutter removal speed, stronger generalization ability, and at the same time marine targets in images are remained completely.

In this letter, a novel fast detection algorithm, known as robust sparse fractional Fourier transform (RSFRFT), is proposed for low-observable maneuvering target detection in a clutter background. The discrete FRFT (DFRFT)-based detection method is time-consuming for large data volumes and the detection performance of sparse FRFT (SFRFT)-based algorithm will be significantly degraded in a heavy clutter background. Using two levels of detection, the defects of DFRFT and SFRFT algorithms are overcome using the proposed algorithm. The first-level detection is performed on the subsampled spectrum to estimate the target frequencies. The second-level detection is carried out after reconstruction for target detection. The simulation analysis and experiments using marine radar data show that the proposed method can achieve a good detection performance for low-observable maneuvering target detection in the clutter background with lower computational complexity.

The micro-motion of a subject induces Doppler frequency modulations around the carrier frequency of the reflected sensor signals. Recently, it has been proved that sea clutter is significantly variable and sea surface subjects have their micro-motions influenced by the sea state. The micro-Doppler (m-D) signatures can describe the refined motion characteristics of sea surface target. The micromotion signal model of sea surface target is established in this paper based on the length of observation time. Finally, the micromotion properties are analyzed using real radar data, i.e., X-band Council for Scientific and Industrial Research (CSIR) data and S-band radar data. It also proves that the m-D can provide extra information of target, which would help improve radar detection and recognition abilities.

Robust and effective detection of a marine target is a challenging task due to the complex sea environment and target's motion. A long-time coherent integration technique is one of the most useful methods for the improvement of radar detection ability, whereas it would easily run into the across range unit (ARU) and Doppler frequency migration (DFM) effects resulting distributed energy in the time and frequency domain. In this paper, the micro-Doppler (m-D) signature of a marine target is employed for detection and modeled as a quadratic frequency-modulated signal. Furthermore, a novel long-time coherent integration method, i.e., Radon-linear canonical ambiguity function (RLCAF), is proposed to detect and estimate the m-D signal without the ARU and DFM effects. The observation values of a micromotion target are first extracted by searching along the moving trajectory. Then these values are carried out with the long-time instantaneous autocorrelation function for reduction of the signal order, and well matched and accumulated in the RLCAF domain using extra three degrees of freedom. It can be verified that the proposed RLCAF can be regarded as a generalization of the popular ambiguity function, fractional Fourier transform, fractional ambiguity function, and Radon-linear canonical transform. Experiments with simulated and real radar data sets indicate that the RLCAF can achieve higher integration gain and detection probability of a marine target in a low signal-to-clutter ratio environment.

In this letter, a novel long-time coherent integration method, known as the Radon-linear canonical transform (RLCT), is proposed for detection of a low observable moving target in sea clutter. The micro-Doppler (m-D) of a sea-surface target is studied and modeled as multiple linear-frequency-modulated signals, which result from the accelerated and 3-D rotated movements. The RLCT-based algorithm employs m-D as a useful signature for target detection and can simultaneously compensate the range and Doppler migrations during long observation time, which simplifies the operational procedure. By searching along the moving trajectory and using extra three degrees of freedom, the observation values of m-D signals can be well matched and accumulated as peaks in the RLCT domain. Then, the target can be declared by comparing the peak value with an adaptive threshold. The definition of the RLCT demonstrates that it is the generalization of the popular moving target detection, Radon-Fourier transform, fractional Fourier transform, and linear canonical transform methods. Finally, experiments using a real sea clutter data set show that the proposed method can achieve high integration gain and detection probability of a micromotion target in heavy sea clutter.

In order to effectively detect moving targets in heavy sea clutter, the micro-Doppler (m-D) effect is studied and an effective algorithm based on short-time fractional Fourier transform (STFRFT) is proposed for target detection and m-D signal extraction. Firstly, the mathematical model of target with micromotion at sea, including translation and rotation movement, is established, which can be approximated as the sum of linear-frequency-modulated signals within a short time. Then, due to the high-power, time-varying, and target-like properties of sea spikes, which may result in poor detection performance, sea spikes are identified and eliminated before target detection to improve signal-to-clutter ratio (SCR). By taking the absolute amplitude of signals in the best STFRFT domain (STFRFD) as the test statistic, and comparing it with the threshold determined by a constant false alarm rate detector, micromotion target can be declared or not. STFRFT with Gaussian window is employed to provide time-frequency distribution of m-D signals, and the instantaneous frequency of each component can be extracted and estimated precisely by STFRFD filtering. In the end, datasets from the intelligent pixel processing radar with HH and VV polarizations are used to verify the validity of this proposed algorithm. Two shore-based experiments are also conducted using an X-band sea search radar and an S-band sea surveillance radar, respectively. The results demonstrate that the proposed method not only achieves high detection probability in a low-SCR environment but also outperforms the short-time Fourier transform-based method.