ABSTRACT Building interior structure sensing using through‐the‐wall radar (TWR) is crucial for safety and security applications. However, existing methods struggle to decouple first‐order wall reflections from second‐order corner interactions, resulting in ambiguity between wall discontinuities and corner features. To address these challenges, this paper proposes a scattering‐order‐based decoupling model integrating geometric optics (GO) and physical optics (PO) principles to enable accurate building characterisation. Our proposed model decomposes total radar returns into distinct components including first‐order scattering from walls and second‐order double‐bounce scattering from corners. For wall characterisation, the scattering response is derived via PO integrals, considering reflections from parallel and perpendicular orientations relative to the radar. For corner detection, a coupled GO‐PO model is developed to explicitly capture double‐bounce interactions between adjacent walls. This analytical formulation isolates the corner representation by excluding first‐order contributions, ensuring the corner signature consists exclusively of coupled scattering energy. As a result, this approach effectively suppresses first‐order wall scattering artifacts and reduces wall‐induced ambiguities. Based on these analytical derivations, we designed a physics‐based dictionary method to enable building interior structure extraction. Extensive simulations and experiments validate that our model improves wall and corner extraction, supporting more reliable building interior reconstruction.

Three-dimensional (3D) building layout sensing is a key capability for Internet of Things (IoT)–enabled smart city applications, including post-disaster assessment and urban security monitoring. However, acquiring reliable 3D building layouts from an external perspective remains challenging in complex urban environments due to severe signal attenuation and multipath effects. This paper proposes an IoT-enabled unmanned aerial vehicle (UAV)–based through-the-wall radar (TWR) sensing framework for large-scale 3D building layout reconstruction. In the proposed framework, UAV-mounted radar sensors act as mobile IoT sensing nodes to collect multi-view sensing data. A multi-layer wall echo propagation model and an angle-weighted three-dimensional back-projection (BP) imaging algorithm are employed to generate multi-view 3D synthetic aperture radar (SAR) representations. An enhanced U-Net architecture is then developed to fuse the multi-view SAR data and reconstruct clearer 3D building layout representations. The simulation and real-world experimental results show that the proposed framework achieves improved reconstruction performance over representative existing methods under the tested conditions, indicating its potential for IoT-oriented smart city sensing.

Building layout sensing of through-the-wall radar (TWR) plays a vital role in fields such as counter-terrorism operations and post-disaster rescue. Existing layout sensing methods based on TWR typically focus solely on either corner information or wall surface features, neglecting the complementarity between the two, which leads to low sensing accuracy in complex environments. To address this issue, we propose a Corner-Wall Sensing Network (CWSNet), a building layout sensing network that fuses corner and wall surface information. First, deep convolutional networks are used to extract wall and corner features from TWR images. Then, these complementary structural features are fused to form an integrated representation. Finally, a transformer-based dynamic graph reasoning module (DGRM) captures their spatial relationships, enabling high-precision layout sensing. Both simulated and real-world experimental datasets demonstrate that CWSNet significantly outperforms existing methods across multiple evaluation metrics, achieving superior wall localization accuracy and layout connectivity, while also exhibiting strong robustness and generalization capabilities.

Through-the-wall radar (TWR) leverages the excellent penetration capabilities of low-frequency electromagnetic waves to achieve noninvasive and nondestructive detection of targets hidden behind the wall. However, the refraction and attenuation phenomena of complex walls have significantly affected the detection performance and imaging quality. Most of the traditional studies have made compensation based on ideal uniform wall structures. They only consider a single medium or a single thickness, which will lead to huge errors for actual nonuniform wall structures. To address the challenges, a multifrequency fusion imaging algorithm that can handle urban nonuniform walls is proposed in this article. First, a segmented wall compensation algorithm is proposed for nonuniform wall scenarios to address poor focusing effects. Second, this article proposes a multifrequency fusion algorithm that integrates high-frequency data with low-frequency data to improve resolution in nonuniform wall environments. Finally, simulation and experimental results demonstrate the superiority of the proposed algorithm under nonuniform wall scenarios.

Noncontact vital sign sensing has attracted increasing attention from both academia and industry. Millimeter-wave radar-based respiration sensing provides high accuracy but suffers from a limited field of view and strong dependence on target position and orientation, which restricts its practical deployment. In contrast, Wi-Fi-based sensing provides wide coverage and great resilience over device setup, yet its respiration estimation accuracy under ideal conditions is generally inferior to that of millimeter-wave radar. Consequently, the robustness of single-modal approaches in practical scenarios remains limited due to the inherent drawbacks of each system. This article presents a fusion-based respiration monitoring system that integrates millimeter-wave radar and Wi-Fi signals. By jointly exploiting the high precision of millimeter-wave radar and the wide-area sensing capability of Wi-Fi, robust respiration monitoring is achieved in practical bedroom environments. To effectively utilize heterogeneous signals, a fusion decision scheme is designed to adaptively determine the necessity of signal fusion. Furthermore, a multimodal signal fusion method based on multivariate signal processing is proposed to jointly extract the common respiration components shared across different modalities. Extensive experimental results demonstrate that the proposed system, FuseRes, can robustly and accurately estimate respiration in complex real-world scenarios, supporting stable noncontact vital sign monitoring and facilitating practical deployment.

Building layout reconstruction (BLR) with through-the-wall radar (TWR) mounted on an unmanned aerial vehicle (UAV) is challenged by clutter and the mismatch between fixed blocks and wall structures. This letter proposes a size-adaptive block-sparse method guided by mean information entropy (MIE). Candidate wall blocks are grown under range-resolution and MIE constraints, and then incorporated into a reweighted alternating direction method of multipliers (ADMM) reconstruction model. To our knowledge, this is the first work to formulate block partitioning for TWR-based BLR as an information-theoretic optimization problem. Simulations achieve the best peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and F1-score among compared methods, reaching 12.511 dB, 0.782, and 0.854, respectively. Field experiments using a UAV-borne linear frequency modulated continuous wave (LFMCW) TWR operating at 2.95 GHz with 500 MHz bandwidth further validate improved structural continuity and clutter suppression.

ABSTRACT With the acceleration of urbanisation, the ability to detect target information in urban blind areas is urgently needed in many fields such as smart cities and autonomous driving. However, existing nonline‐of‐sight (NLOS) target localisation methods mainly assume that the urban building walls are flat with ideal reflections. This simplification results in significant localisation errors in real‐world scenarios, where wall surfaces include concave–convex features such as balconies and windows. To address this issue, we propose a joint estimation algorithm for target position and wall parameters based on the single‐input‐single‐output (SISO) radar system. First, we parameterise the concave–convex features of the walls at reflection points and incorporate them into the multipath echo model based on the sparsity of the scene. Then, an objective function is constructed by minimising the error between the echo signal and the signal reconstructed using compressed sensing. Furthermore, the alternating direction method of multipliers (ADMM) and genetic algorithms are designed for hierarchical optimisation. Finally, extensive simulations and experiments demonstrate that the proposed algorithm can achieve high‐precision NLOS target localisation in irregular wall surfaces.

Material identification based on intelligent sensing is a fundamental capability that enables next-generation IoT applications such as industrial automation, smart-home perception, and safety detection. Existing technologies often suffer from limited deployment flexibility, poor robustness when dealing with small objects, or insufficient exploitation of material physical properties. To address these issues, this paper proposes a non-contact material identification method based on millimeter-wave radar and micro-vibration sensing. By acoustically exciting micro-vibrations on the target and extracting micro-vibration features from the radar echoes, the method effectively distinguishes different material types. First, we design a same-side excitation detection architecture, in which the acoustic source and radar are placed on the same side of the target. This configuration enhances practical deployment flexibility and mitigates the issue that, under opposite-side setups, the weak echoes from small targets may be obscured by strong reflections from the speaker. Second, we extract three types of micro-vibration features from the radar echoes—frequency, power, and damping—each capturing distinct physical aspects of material properties. Furthermore, to effectively fuse these heterogeneous features, we develop a feature-aware neural network that employs branched modeling and attention mechanisms based on the structural characteristics of different features. The experiments are conducted on eight common material categories, each with multiple object instances and measured under various target-radar distances and observation angles. The results demonstrate that the proposed method achieves an average identification accuracy of 99.0%, providing a practical solution for non-contact material identification in IoT scenarios.

Real-time bed-exit detection for the elderly or bedridden patients serves as a pivotal strategy in healthcare to mitigate accident risks and improve nursing standards. To address the challenges posed by covered stationary human targets and dynamic environmental interference, this paper proposes a novel non-contact detection method based on millimeter-wave (mmWave) radar. First, we introduce an adaptive clutter suppression mechanism. We fuse global energy with R´enyi entropy to filter out interference from fluttering light fabric. Second, to detect weak targets and filter out isolated noise clusters, we design a cascaded framework comprising ordered-statistic constant false alarm rate (OS-CFAR), adaptive morphological filtering and area thresholding. Finally, to suppress interference from fluttering heavy fabric, we propose a spatiotemporal morphological consistency assessment. Experimental validation based on 14.5 h of data collected from real-world scenarios demonstrates that the system achieves a sensitivity of 99.0%, a specificity of 98.2%, and an average detection delay of 6.9 s. These results confirm the reliability and robustness of the proposed method in handling complex scenarios, including blanket occlusion and interference from fluttering bedding and curtains.

Through-wall radar (TWR) plays a pivotal role in nonintrusive detection of buildings because of its excellent penetrability. Many methods have been developed to estimate the parameters of walls and perform wall compensation for accurate through-wall imaging results. However, most existing studies are based on uniform wall models that cannot be generalized to actual building walls with complex structures such as doors, pillars, and windows, nor can they quickly generate a parameter distribution map of the entire wall. To solve this problem, this article proposes an innovative segment-then-estimate framework that first fuses LiDAR point clouds and optical images with simultaneous localization and mapping (SLAM) to reconstruct a 3-D exterior wall model, thereby guiding the precise division of radar brightness scan (B-scan) data along the slow-time dimension. Then, we introduce a Transformer-based parameter estimation model for the echo data of each segment, leveraging its self-attention mechanism to capture long-range dependencies and deep features to accurately estimate the wall thickness and relative permittivity. Finally, segment-based estimates are stitched using geometric information obtained from multimodal reconstruction to generate the wall parameter distribution map. Experimental results demonstrate that the wall parameter estimation maps generated by the proposed framework in this study not only achieve highly accurate parameter estimation values but also exhibit distinct door and window boundaries, providing reliable prior information for highly refined compensated imaging.

Through-the-wall radar is a vital tool to sense the indoor scene, while multiple reflections of electromagnetic waves in the room produce multipath ghost images, resulting in false alarms in target detection. The location of the ghost image is determined by the target position, indoor layout, and array structure. Accordingly, we induce changes in the ghost image's appearance by applying phase compensation to simulate the geometric delay of a rotated array, which requires no prior information about the indoor layout or neural network training. Concretely, the proposed method verifies and extracts aspect dependence by performing phase compensation signal processing of the array center rotation on the echo data and then achieves multipath ghost suppression through cross-correlation back projection and incoherent fusion. The results of simulation experiments and real data experiments demonstrate the effectiveness of the proposed method.

Through-the-wall building interior structure sensing has been greatly serving in various applications, including search-and-rescue operations. However, most existing methods exhibit limitations in imaging the walls and corners with good continuity and recognizable features. In this article, we consider imaging of the building interior structures by extracting the major building elements with structural continuity. Specifically, the signals from a complex building are first modeled as the superposition responses from discrete canonical scatterers, such as planar walls and wall corners. Then, a structural variational Bayesian method is designed to detect and extract these critical structures. This method improves the 1-D continuity of the walls and the 2-D continuity of the corners through a Bayesian hierarchical probabilistic model. Moreover, we incorporate the generalized approximate message-passing technique into the variational expectation maximization method to efficiently estimate the walls and corners simultaneously. Results from both simulated and real data validate the effectiveness of the proposed method in accurately extracting walls and corners with improved continuity, thereby enabling a comprehensive building structure.

With the acceleration of urbanization, through-wall radar (TWR) technology has become crucial for military reconnaissance and disaster emergency response. However, conventional TWR systems predominantly adopt oversimplified homogeneous wall models in their compensation algorithms, which fail to account for the inherent heterogeneity of real-world building walls. To mitigate the impact of irregularities such as doors, windows, pillars, and protrusions on nonuniform building walls during through-wall imaging, this article proposes a multisource fusion system and a wall compensation method based on light detection and ranging (LiDAR) point cloud data. By integrating measurements from both LiDAR and TWR, and employing simultaneous localization and mapping (SLAM) technology along with point cloud preprocessing and contour fitting techniques, the system generates accurate wall contour information, enabling effective compensation. The experimental results show that the accuracy of building interior layout reconstruction is significantly improved by compensating for the effects of external wall irregularities extracted from LiDAR data.

The classical constant false alarm rate (CFAR) detector is optimal for target detection in Gaussian white noise but struggles with unknown, time-varying sea states. Data-driven target detection methods are highly sensitive to clutter distribution, leading to poor detection performance and high false alarm rates ( P_fa ) under unknown sea states. This paper proposes a deep feature constant false alarm ratio (DF-CFAR) detector based on a feature game model. The input to the feature extraction network is fragmented fast time-dimensional data processed by coherent integration, incorporating a feature game mechanism. This mechanism effectively mines target echo features that are uncorrelated with the background sea clutter. By transforming the target detection problem under varying sea state conditions into a regression prediction problem, we achieve constant false alarm detection through a statistical probability threshold. Simulation and real data experiments demonstrate that, compared to commonly used algorithms, the proposed method not only achieves higher detection probability and lower false alarm rates but also exhibits excellent constant false alarm characteristics and stronger robustness to unseen sea states. Results from publicly available X-band radar data processing show that the proposed algorithm can eliminate the influence of different sea clutters and improve the detection performance of sea-surface small targets by an equivalent signal-to-clutter ratio (SCR) of about 3.1 dB.

Through-wall sensing (TWS) systems have extensive applications in civilian as well as military fields because of its ability to detect the obscured space behind obstacles. To harvest good penetrability at a low cost, existing TWS systems often use sparse array configuration operating in the L/S electromagnetic wave band, which poses a poor spatial resolution in the radar images. As a result, it is difficult for end-users to identify a target from the through-wall radar images because the lack of geometry-related information such as outline, shape, etc. In response to this challenge, this article proposes a high-resolution TWS imaging method by the conditional denoising diffusion probabilistic model (DDPM). First, we design a hybrid encoder to extract and fuse the feature from multisource data including the 3-D radar images and 2-D optical images. The extracted features are fed into the network consisting of residual and self-attention modules to predict/estimate the noise, which is then subtracted from the current image. Finally, by estimating and subtracting the noise iteratively, we can obtain the high-resolution image. Simulations and real-world experiments confirm the efficiency of the proposed method in successfully reconstructing the outline and contour information of the target, which outperforms most existing TWS systems in resolution aspect.

Objective Small targets such as unmanned aerial vehicles and unmanned vessels, which exhibit small Radar Cross Section (RCS) values and weak echoes, are difficult to detect due to their low observability. Traditional Constant False Alarm Rate (CFAR) detection is typically represented by the Cell-Averaged (CA) CFAR method, in which the detection threshold is determined by the statistical power parameter of the signal. However, its detection performance is constrained by the Signal-to-Noise Ratio (SNR). This study focuses on how to exploit and apply signal features beyond power parameters to achieve CFAR detection under lower SNR conditions. Methods After pulse compression, the envelope of a Linear Frequency Modulation (LFM) signal exhibits sinc characteristics, whereas noise retains its random nature. This difference can be used to distinguish target echoes from non-target signals. On this basis, we propose a constant false alarm detection method based on signal feature matching. First, both the ideal echo signal and the actual echo signal are processed with sliding windows of equal length to generate an ideal sample and a set of test samples. A dual-port fully connected neural network is then constructed to extract the deep feature matching degree between the ideal sample and the test samples. Finally, the constant false alarm threshold is obtained by numerically calculating the deep feature matching parameter from a large number of non-target samples compared with the standard sample. Results and Discussions Several sets of simulation experiments are carried out, and measured radar data from different frequency bands are applied to verify the effectiveness of the proposed method. The simulations first confirm that the method maintains stable constant false alarm characteristics (Table 1). The detection performance is then compared with traditional CA-CFAR detection, machine learning approaches, and other deep learning methods. The results indicate that, relative to CA-CFAR detection, the proposed method achieve 2'5 dB gain in equivalent SNR across different false alarm probabilities ( Fig. 4). Under mismatched SNR conditions, the method continues to demonstrate robust detection performance with strong generalization capability (Fig. 5). In the processing of measured X-band radar data, the proposed method detects targets that CA-CFAR fails to identify, extending the detection range to 740 distance units, compared with 562 distance units for CA-CFAR, corresponding to an improvement of approximately 28.72% in radar detection capability (Fig. 7, 8). In the case of S-band radar data, the proposed method significantly reduces false alarms (Fig. 10, 11). Conclusions This study exploits the difference between target and noise signal envelopes by introducing a feature extraction network that effectively enhances target detection performance. Comparative simulation experiments and the processing of measured radar data across different frequency bands demonstrate the following: (1) the proposed method markedly improves detection performance over traditional CA-CFAR detection, yielding a 2'5 dB gain in equivalent SNR; (2) under mismatched SNR conditions, the method shows strong generalization capability, achieving better detection performance than other deep learning and machine learning approaches; (3) in X-band radar data processing, the method increases detection capability by approximately 28.72%; and (4) in S-band radar data processing, it significantly reduces false alarms. Future work will focus on accelerating the detection process to further improve efficiency.

Through-wall radar (TWR) imaging is broadly applied in the detection of enclosed space, which plays an important role in security, rescue, and military operations. However, the resolution of the existing TWR imaging method is not high enough to serve the practical application because of the physical constraints, such as the limited antenna aperture. The deep learning method can improve the resolution due to its powerful reasoning capabilities, but it is challenging and costly to build comprehensive TWR datasets for training in practice. To address these challenges, this article proposes a less-forgetting high-resolution network (LFHRNet) for TWR imaging. First, a cGAN network is trained on the source data as the pretrained network to initialize the target network. The target network LFHRNet integrates two parallel branches by a router network. It forces a branch to focus on learning knowledge to solve target tasks, while another is frozen to alleviate the catastrophic forgetting. Finally, during the online phase, the low-resolution radar image in both the source and target domains can be input into LFHRNet to get the high-resolution image. Simulation and practical experimental results show that LFHRNet can learn knowledge to reconstruct the shapes of the objects in the target domain, while maintaining the source knowledge to reconstruct the shapes of the objects in the source domain.