Iet Radar Sonar and Navigation最新文献

Due to the stable propagation of magnetic signals in ocean and air, magnetic detection technology has become an effective means for nonacoustic detection. The magnetic anomaly signal and shaft-rate magnetic signal radiated by underwater vehicles are currently the most effective magnetic detection signals. Existing magnetic detection methods primarily focus on studying either magnetic anomaly signal or shaft-rate magnetic signal. However, since a target can generate both of these magnetic signals simultaneously, detecting one type may lead to the neglect of the other, reducing detection accuracy. To overcome the limitations of existing technologies, this paper presents a combined detection method for magnetic anomaly and shaft-rate magnetic signals. The detection process is divided into magnetic anomaly signal detection based on orthogonal basis function (OBF) and shaft-rate magnetic signal detection based on adaptive line spectrum enhancement (ALE). Especially for the detection of magnetic anomaly signal, this paper proposes a preprocessing method based on the LOESS smoothing technique, utilising noise characteristics, and combines it with the CFAR criterion for decision-making. This approach significantly improves the detection accuracy of the magnetic anomaly signal. Finally, the simulation and experimental results show that combining magnetic anomaly and shaft-rate magnetic signals for combined detection can effectively improve the detection accuracy.

Sea clutter suppression is a hot topic for airborne radar. Space-time adaptive processing (STAP) is a useful approach to address this issue. Currently, range ambiguity is a problem to restrict conventional STAP performance. The conventional STAP cannot differentiate the signals originating from distinct ambiguous areas due to the lack of range-associated degrees-of-freedom (DoFs). Frequency diverse array (FDA) can provide the DoFs via introducing a frequency shifting between adjacent transmit elements, then FDA-STAP is developed. According to the Reed, Mallet and Brennan (RMB) rule, FDA-STAP requires more training samples in comparison with conventional STAP. However, the number of training samples is limited practically, and FDA-STAP will suffer from severe performance deterioration. To address this issue, this paper introduces a sparsity recovery algorithm, atomic norm minimisation (ANM), into FDA-STAP for clutter profile recovery, that is, ANM-FDA-STAP, thereby reducing the requirement on training samples. Numerical results verify that the ANM-FDA-STAP algorithm exhibit outstanding performance.

This work addresses the problem of multiplatform motion error calibration in an asynchronous airborne radar network with the employment of ground strong scatterers (SSs) at inaccurate locations. The multiradar time-frequency synchronisation errors as well as motion errors, including position and velocity errors, would significantly impair signal coherence of airborne radar networks, provided that a potential target can share the same complex scattering coefficient to all airborne radars distributed within hundreds of yards. Hence, the network configuration calibration under time and frequency asynchronisation is required prior to coherent beamforming. Starting from the nonlinear time of arrival (TOA) and frequency of arrival (FOA) measurement equations, we develop an iterative reweighted least squares (IRLS) algorithm to obtain the deviations in positions, velocities, instants and frequencies of multiple radars during each iteration. By adding the deviations obtained from all iterations, a final error estimation is achieved and the evaluation of multiradar parameters is more refined. During algorithm development, we apply Taylor series expansion to eliminate nuisance parameters, followed by reweighted iterations to manage the remaining nonlinearity. This approach allows us to form linear equations for estimating multiradar parameter errors. Besides, we conduct the performance analysis of our method in comparison with the theoretical Cramér-Rao lower bound (CRLB). Both theoretical derivations and simulation results confirm the effectiveness of our algorithm.

This study analyses sonar performance for underwater object detection in four regions of the Marmara Sea, using oceanographic data from the Turkish Naval Forces and open source datasets. Simulations were conducted with LYBIN acoustic modelling software across four seasons (January, May, July and October), evaluating variable-depth sonar (VDS) and hull-mounted sonar (HMS) systems for coverage and detection performance. Results identified optimal sonar coverage zones, highlighting seasonal impacts on propagation, with temperature and salinity fluctuations directly influencing performance. Seasonal stratification in the Marmara Sea generates surface ducts and shadow zones that strongly constrain HMS performance, while VDS consistently mitigates these effects. Simulations demonstrate that VDS reduces shadowed areas by 25% across all seasons and regions, extending reliable detection ranges compared with HMS. The study provides a foundation for designing efficient underwater surveillance systems in the Marmara Sea, offering insights for optimising operational strategies. Future research should explore diverse marine conditions and sonar configurations to enhance detection capabilities.

Mechanically generated sounds, common in industrial process control and surveillance, often exhibit narrowband harmonic features that manifest as line spectra in the time–frequency domain. While convolutional neural networks (CNNs) have been employed for line spectrum extraction, their performance is often hindered by the scarcity of high-quality supervised data. To address this limitation, we explore graph neural networks (GNNs), which explicitly model feature relationships. Among GNNs, graph convolutional networks (GCNs) stand out due to their computational efficiency. In this study, we introduce a GCN model enhanced with a weight tensor to effectively extract line spectral features from graph representations of mechanical sounds. Our approach is tailored for weakly supervised scenarios, where time–frequency masks are noisy and interfere with supervision. By leveraging a tensor product operation, the model projects input graphs into a multi-dimensional embedding space, facilitating the learning of diverse and discriminative representations with minimal computational overhead. Experimental results on audio and underwater acoustic datasets reveal that our method outperforms fully supervised baselines while significantly reducing computational requirements. These results underscore the efficiency and practicality of our framework for real-world acoustic processing applications.

The resonance parameters of late-time returns (LTRs) can be used as features in the identification of radar targets. However, reliable estimation of the complex frequency of each resonance is notoriously difficult. This is a result of the short duration of the LTR, its low effective signal-to-noise ratio (SNR) and the inherent sensitivity of the estimation problem. These issues are exacerbated when the radar background includes resonating clutter. We develop an effective technique for estimation the complex frequencies of a target's resonances for scenarios in which the radar can obtain multiple measurement shots of the background (clutter) alone and multiple measurement shots of the target in the presence of the background. The proposed method exploits the fact that the maximum likelihood estimator for measurements in Gaussian noise can be decomposed to estimate the complex frequencies of the resonances separately from their complex amplitudes. This enables us to decouple the estimation of the complex frequencies of the target from those of the background because the background's complex frequencies remain largely unchanged when the target is introduced. We investigate the performance of the proposed method using a radar that operates in the band of 0.5–5 GHz and employs equivalent sampling at a rate of 20 GSa/s. Proof-of-concept experiments on brass rods of known length validate the overall approach, and experiments on more complex targets in clutter demonstrate its potential for practical applications.

Current radar signal deinterleaving methods primarily utilise direction of arrival, radio frequency and pulse repetition interval to separate signals from different radars. However, their effectiveness becomes limited in scenarios with highly overlapping radar parameters. Amplitude can provide supplementary discrimination for many deinterleaving problems, especially for mechanically scanning radars. The amplitude of signals intercepted from such radars exhibits a continuous parabolic-like variation characteristic. Leveraging this, we construct a function-adapted Gaussian mixture model to characterise the joint distribution of pulse time-of-arrival and amplitude for interleaved pulse trains, thereby transforming radar signal deinterleaving into a parameter estimation and clustering problem. Furthermore, we employ active function cross-entropy clustering (afCEC) to solve the problem and innovatively embeds the sequential andom sampling consensus within the afCEC framework to mitigate its sensitivity to initial values and avoid local optima. This achieves preliminary clustering of the time-amplitude data, effectively decomposing the originally pulse point cloud into multiple subclusters conforming to mixture model components. Building upon this over-segmentation result, we design a merging strategy based on pulse cluster continuity, enabling automatic deinterleaving without prior knowledge of radar quantity. Simulation results demonstrate that the proposed method achieves superior deinterleaving performance in complex electromagnetic scenarios, outperforming state-of-the-art approaches.

This paper proposes a dual-backbone feature fusion approach to address the few-shot class imbalance problem in specific emitter identification. First, employ the Weighted Random Sampler algorithm to dynamically calculate sampling weights for data preprocessing; Subsequently, by fusing the two single-backbone networks of ResNet50 and ConvNeXt-Tiny, we overcome the hierarchical limitations of their independent feature capture, thereby achieving few-shot multi-scale and multi-level feature extraction while enhancing fine-grained features; Furthermore, we embed Efficient Channel Attention into the dual-backbone networks to achieve dynamic modelling of inter-channel correlations. This method enhances feature attention on ‘minority class’ samples while suppressing redundant information, thereby improving the accuracy, stability and robustness of specific emitter identification under imbalanced data conditions. Experimental results validated on a public Bluetooth dataset demonstrate that the proposed method achieves at least a 6% improvement in recognition rate compared to other commonly used algorithms.

The line spectrum from ship-radiated noise is a critical feature for passive sonar to detect underwater acoustic targets. However, due to weak target strength as well as severe propagation attenuation and oceanic ambient noise, the signals received by passive sonars generally manifest low signal-to-noise ratio (SNR), strong nonstationarity and overwhelmed Doppler-shifted line spectrum. These challenges deteriorate the performance of conventional Cohen class time frequency distribution (CCTFD) methods in capturing the fine spectral feature of such signals. To overcome these difficulties, this research proposes an improved Cohen-class method, termed ambiguity function-instantaneous autocorrelation function joint filtering Wigner–Ville distribution (AIJF-WVD). First, this study analyses how standard CCTFD's cross-term suppression mechanism degrades time-frequency resolution/concentration in time-frequency distribution (TFD) when processing multicomponent Doppler-shifted signals. Departing from conventional framework of cross-term suppression via two-dimensional low-pass filtering along both frequency-shift dimension and time-delay dimension in ambiguity function (AF) domain, AIJF-WVD presents a novel joint filtering approach that consists of designing one-dimensional finite impulse response (FIR) filter solely along frequency-shift dimension in AF domain (while maintaining time-delay dimension unchanged) as well as subsequent one-dimensional low-pass filtering along time dimension in instantaneous autocorrelation function (IAF) domain based on the designed filter. Therefore, this novel method enhances TFD performance of cross-term suppression and frequency resolution simultaneously while maintaining low computational complexity. Then, the performances of various CCTFD methods are quantitatively assessed using mean structural similarity (MSSIM), normalised Rényi entropy (NRE), half-power bandwidth (HBW) and mean runtime. Finally, the global spectral estimation accuracy of Doppler-shifted tonals is evaluated through tracking deviation analysis. Compared to conventional CCTFDs, AIJF-WVD exhibits superior robustness and adaptability in low-SNR background noise as evidenced by processing both simulated signals and ship-radiated noise from sea trials. Furthermore, the refined approach is also validated to significantly improve cross-term suppression, time-frequency concentration and computational efficiency characteristics while preserving frequency resolution and superior tonal trajectory tracking capability for passive sonar.

This paper presents a novel approach to synthesising radar echoes for unmanned aerial vehicle (UAV) anticollision systems, specifically focusing on the challenges posed by high-rise architectural landmarks in urban environments. We employ a Monte Carlo method to generate synthetic radar data that accurately reflects the statistical properties of real-world radar echoes, derived from data collected using a custom-designed X-band radar. Our methodology involves the probabilistic modelling of radar echoes for three distinct classes: large-scale arena building, sky-scraping slender spire and background noise, using kernel density estimation (KDE). This approach allows for the creation of a large database of synthetic radar signatures essential for training and validating machine learning algorithms intended for use in UAV collision avoidance systems. The results demonstrate the efficacy of our method in preserving the statistical characteristics of real radar data, enabling the generation of high-fidelity synthetic echoes that can significantly enhance the development and testing of UAV navigation and obstacle avoidance systems.