Iet Radar Sonar and Navigation最新文献

Unmanned aerial vehicles (UAVs) heavily rely on GPS, a system vulnerable to signal interference in complex urban environments. Although radar systems offer a robust alternative due to their ability to effectively penetrate adverse weather and operate in darkness, a key challenge remains: reliably identifying static architectural landmarks from sparse and noisy radar echoes. This paper proposes a novel method for creating spatio-probabilistic models (SPMs) of radar echoes from high-rise urban landmarks, enabling independent, radar-based UAV localisation. We employ kernel density estimation on real radar data, acquired with a custom-designed X-band ENAVI radar, focusing on large arena buildings and slender spires. These SPMs are then used to detect and identify landmarks by calculating the divergence between the probability distributions of the real-time received echoes and the preestimated reference models. Our evaluation, using probabilistic divergence metrics on Wrocław's Centennial Hall and Iglica, shows that this method effectively preserves the statistical properties of the radar data, generating high-fidelity SPMs. This approach significantly improves landmark identification compared to classical correlation methods, paving the way for more robust and resilient UAV navigation systems.

Nano-drones are insect-like drones used to provide intelligence through their capability of intrusion and ability to carry small sensors. They pose a defence and security threat and can potentially violate secure establishments and privacy rights. Their rapid emergence and increased availability have made them an existing technology which is affordable and easy to operate. Nano-drones are typically defined as drones smaller than 15 cm. They are light and stealthy in nature and present a very low radar cross-section (RCS) which creates a significant challenge for active Radio Frequency (RF) security systems tasked with detecting potential threats. This paper presents a K-band Frequency Modulated Continuous Wave (FMCW) radar prototype tailored for detecting nano-drones. Operating at 24 GHz and utilising commercial off-the-shelf components, the radar offers a low-cost, flexible and customisable solution with user-selectable frequency and waveform parameters. The system's detection capabilities were tested using low-RCS oscillating metallic spheres ranging from 0.5 to 3.0 cm in diameter. Nano-drone detection was demonstrated using range-Doppler maps and time-frequency spectrograms of a real and small 5 cm nano-drone. The paper provides a detailed overview of the radar design and test methodology, together with a detailed investigation of the radar performance.

To address the issues of poor detection performance under low signal-to-noise ratios (SNRs) and high computational complexity in existing visibility graph-based spectrum sensing algorithms, this article proposes a novel algorithm based on the Euclidean norm of the horizontal visibility graph (HVG) adjacency matrix. The algorithm begins by computing the block summation of the observed signal's power spectrum. The squared modulus of its autocorrelation function is subsequently calculated, normalised and quantised to form the new sequence, which is then transformed to the HVG and defined as the graph signal. The one-hop graph filter is constructed from the graph signal and the adjacency matrix, and its Euclidean norm serves as the detection statistic. This statistic is compared against a predefined threshold to determine the presence of the primary user signal. To theoretically analyse detection performance, the weak submajorisation order is introduced to evaluate the statistical differences between graph signals under the two hypotheses. Additionally, data exploration demonstrates that the proposed statistic approximately follows a Burr distribution under the null hypothesis, allowing for an approximate analytical expression for the detection threshold is derived. Simulation results show that the proposed algorithm outperforms existing graph-based algorithms at low SNRs while maintaining moderate computational complexity.

In some ballistic target tracking applications, the target travels to the destination with a constant horizontal heading in the reentry phase, whose states are subjected to a destination constraint. If the prior information on the destination can be acquired and effectively utilised, a significant enhancement of performance can be expected. In this paper, a three-dimensional (3D) constrained motion model is established to describe the target motion in the reentry phase. For different cases where the prior destination information is accurately known or contaminated by noise, the horizontal heading angle or the destination position is augmented into the state vector to formulate the accurate constraint relationships in the horizontal plane. Based on the augmented state vectors and the existing 2D model for reentry targets in the vertical plane, accurate state equations are derived to describe the ballistic target motion in the 3D space. Corresponding filtering methods, which employ the unscented Kalman filter to deal with the strong nonlinearity in the augmented state equation, are proposed. Simulation results of Monte Carlo experiments verify the effectiveness of the proposed constrained estimation methods. It is demonstrated that the incorporation of extra destination constraint information leads to superior tracking performance compared with the unconstrained method.

This study presents a deep neural network (DNN) model for multi-octave-band direction-finding (MOB-DF) estimation using a broadband DF-array and multi-layer perceptron (MLP). The model leverages randomly placed array elements to generate unique array steering vectors (ASVs) for directions within a cone-shaped field-of-view. By directly linking ASVs and signal frequency to direction via an MLP, it eliminates reliance on the signal covariance matrix, a common component in many 2D neural network-based DF methods. The DNN-based MOB-DF model is structured into sub-bands, each utilising a trained 16 × 1024 MLP. Simulations with 3-, 4-, and 5-element DF models, trained and validated on datasets with signal-to-noise ratios (SNRs) of 10, 20, and 100 dB respectively, reveal several key findings: (1) MLPs trained at 10 dB SNR can achieve better estimation performance across varying SNR levels, where estimation performance is defined as the probability of direction estimation error ≤ 1°. (2) Increasing array elements expands MOB coverage. (3) The 5-element model attains probabilities of 50% and 90% for ≤ 1° estimation errors at approximately −20 and −10 dB SNR respectively within 2–20 GHz. (4) Average prediction time per direction is on the microsecond scale. (5) The model shows resilience to frequency estimation uncertainties.

Conventional Loran-C is mainly used for low-altitude users; however, when the Loran-C signal station or receiving point is at a higher altitude, the ranging error caused by the elevation change cannot be ignored. The traditional groundwave path correction method for high altitude regions idealises the complex groundwave path as a smooth, extensive elliptic line. However, this is a rough and inaccurate correction value $��\left(��\Delta �S��\right)�$ for the groundwave path. In this paper, based on the Huygens–Fresnel principle, we analyse the Loran-C groundwave path, and propose the Groundwave path accumulation (GPA) method, which calculates the complex terrain groundwave transmission paths in segments, to solve the problem of the low accuracy of $��\Delta �S�$ in the traditional method. With the opening of high-altitude Loran-C stations in western China, the algorithm in this paper can improve the accuracy of Loran-C users' packet positioning to a certain extent in central and western China, Central Asia, and South Asia. The article analyses the correction value of the GPA algorithm to the Loran-C ground wave transmission distance between two points with elevation, and the ground wave path correction value is 46.918 m in the elevation difference of 2500.000 m and no elevation distance of 414,306.538 m.

Degrees of freedom (DOF) serves as a critical metric for evaluating the design of sparse arrays. Developing novel sparse arrays with enhanced degrees of freedom and mathematically expressible structures constitutes a significant research direction in the field of direction of arrival (DOA) estimation. In this paper, an innovative V-shaped sparse sensor array is proposed through the strategic adjustment of sensor positions within the array which is called V-shaped fourth-order linear array (VFLA). Compared to existing V-shaped sparse arrays, the proposed configuration demonstrates superior degrees of freedom when exploiting the covariance of received signals for DOA estimation. Furthermore, relative to commonly used V-shaped arrays and other sparse arrays, the VFLA exhibits not only higher degrees of freedom but also a larger array aperture, thereby enhancing the accuracy of two-dimensional (2-D) DOA estimation. Finally, simulation experiments validate the outstanding performance of the VFLA in 2-D DOA estimation.

Recent developments in driving technology have led to the creation of advanced driver assistance systems and progress towards fully autonomous vehicles. Cars equipped with radar technology can simultaneously detect multiple vulnerable road users, assessing their distance, speed, and approach angle. For autonomous vehicles to be deemed safe for public roads, they must effectively identify and classify these users. This study employs time–frequency analysis and deep learning techniques to classify spectrograms derived from targets. The training and testing datasets were generated using frequency-modulated continuous-wave (FMCW) radar signals operating at 77 GHz. A five-layer convolutional neural network (CNN) was trained for this purpose. We investigated how different time window types and durations affect the Short-Time Fourier Transform calculation and the CNN classification accuracy for each scenario. As the length of the time window increases, frequency resolution improves, enabling better differentiation between closely spaced frequencies and enhancing classification accuracy. However, increased time window lengths lead to decreased time resolution, causing accuracy to plateau at 800; beyond this point, accuracy declines. We achieved an accuracy rate of 88.95% in classifying seven data classes, with improvements in specific classes compared to prior studies. The findings suggest that micro-Doppler-based convolutional neural networks can effectively classify vulnerable road users, contributing to collision avoidance efforts.

This article discusses the problem of tracking a single 3D extended target (or widely separated targets) with convex polytope shape when the target may only be partially visible. An extended target (as opposed to a point target) may generate multiple measurements in a single frame. With the advent of high-resolution sensors (such as LiDAR), the targets need to be considered as extended targets and their shape as well as kinematics need to be estimated. The extended target may only be partially visible (self-occlusion) and the measurements occur only from the visible parts of the target. In this work, different parts of a single extended target are assumed to be different targets constrained by the rigid body motion of the whole target, and the multitarget tracking framework is used to handle the tracking. The target shape is described using a convex hull represented by its vertices and a Delaunay triangulation. The point target PMHT is modified to develop an extended target PMHT (ET-PMHT) joint association and filtering by assuming that the face triangulations are separate targets. Face management is incorporated into the algorithm to delete erroneous faces and the algorithm is able to add new faces to refine the shape estimate. The framework can handle self-occlusion (partial visibility) by associating measurements only to the visible parts of the target. The algorithm's performance is compared with the 3D Gaussian Process under various scenarios, and RMSE of the centre, velocity and IoU metrics are used to quantify the performance. The proposed algorithm is able to outperform the 3D Gaussian Process in the centre RMSE metric by about 40% while achieving an IoU of 0.6 (on average) even when the target is only partially visible.

Image forgery detection aims to identify tampered content and localise manipulated regions within images. With the rise of advanced editing tools, forgeries pose serious challenges across media, law and scientific domains. Existing CNN-based models struggle to detect subtle manipulations that mimic natural image patterns. To address this challenge, we propose a dual-stream contrastive learning network (DSCL-Net) that jointly exploits spatial (pixel-level) and frequency (noise-level) cues. The architecture employs two ResNet-50 encoders: one processes the red–green–blue (RGB) image to capture semantic context, whereas the other processes a spatial rich model (SRM) filtered version to extract high-frequency forensic traces. A multi-scale attention fusion module enhances manipulation-sensitive features. The network includes three heads: a classification head for image-level prediction, a segmentation head for pixel-wise localisation, and a contrastive projection head to improve feature discrimination. We validate the effectiveness of our proposed model on two benchmark datasets. The proposed DSCL-Net surpasses previous state-of-the-art methods by achieving an image-level accuracy of 97.9% on the CASIA and 89.8% on IMD2020. At the pixel level, it attains an F1-score of 92.7% and an AUC of 91.2% on CASIA, and an F1-score of 86.6% with an AUC of 90.1% on IMD2020. Furthermore, LIME and SHAP have been employed to provide explainability at individual image level to comprehend the alignment of the predicted mask with the ground truth mask. The developed approach contributes to the development of safe technology for dealing with misinformation and fake news.