JGJunbo Gong

Junbo Gong

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Junbo Gong has a total of 38 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Intelligent classification of bonding status for building external thermal insulation layers using signal-driven multi-module fusion with ground-penetrating radar

    by Yuhan Li, Changjun Wang, Zihang Jiang, Sen Pang, Junbo Gong, Shuangying Sun, Jing Wen, Kunao Li, You Li, Tian Lan

    Ground-penetrating radar (GPR) serves as a critical method for the non-destructive evaluation of bonding status in building external wall thermal insulation system. However, existing approaches predominantly differentiate bonding status based on visual features, resulting in low detection accuracy and poor interpretability for debonding defects with highly similar morphologies. Additionally, affected by measurement line shifts, single-slice echo signals exhibit weakness and incomplete feature representation, which limits generalization capabilities in small-sample scenarios. To address these limitations, this study proposes an end-to-end intelligent detection framework with multi-module integration. Firstly, a signal-visual dual-stream network is constructed, where cross-modal attention enables dynamic guidance of visual features by radar signals, generating sparse candidate regions to compensate for missed detections in weak signals. Secondly, a deformable cross-slice fusion network is designed to aggregate multi-line information through self-supervised pseudo-target selection and pixel-level spatial correction, thereby enhancing weak echo features. Thirdly, a hybrid feature classification framework is established by integrating handcrafted features with deep learning features, combined with transfer learning of pretrained weights and stratified sampling Extreme Gradient Boosting (XGBoost), achieving precise classification of three bonding status. Finally, experimental validation on measured datasets demonstrates an overall classification accuracy of 92\%, maintaining superior performance even for small-sample debonding samples, thus confirming its effectiveness in complex engineering scenarios. This research provides reliable intelligent technical support for non-destructive defect detection in building external thermal insulation systems and holds significant engineering implications for advancing smart non-destructive testing technologies in industrial infrastructure.

  • Robust and Lightweight 3-D Reconstruction of Buried Threats Using Multiview GPR Data

    by Shiwen Sheng, Xiaopeng Yang, Weicheng Gao, Zexi Wang, Junbo Gong, Tian Lan

    Terrorist attacks pose a severe threat to global public security, among which preembedding offensive weapons in walls is an important attack type. Therefore, obtaining information about buried objects is crucial for addressing potential threats. Ground-penetrating radar (GPR), as an established nonintrusive detection method, is capable of effectively acquiring the 3-D information of buried objects. However, traditional GPR imaging methods often fail to complete the 3-D reconstruction task for targets under conditions of low signal-to-noise ratio (SNR). Moreover, the large computational load limits its application in real-time detection scenarios. To address these challenges, this article proposes a robust and lightweight 3-D reconstruction method for buried targets, specifically for pistols, eavesdropping devices, and explosives. The method first utilizes the 3-D Fourier transform of the wave equation to analyze the target wavefield information from 3-D C-scan data, which can obtain a preliminary visualization of the target, and then employs a projection mechanism to process the sparse energy-focused volume from multiple views. Finally, a multiview reconstruction algorithm is used to reconstruct the 3-D voxel model of the target. Compared to existing methods, our approach reduces parameters by 98.9% to 48.59M versus 3-D U-Net, floating-point operations (FLOPs) by 79.7% versus Kirchhoff migration, and achieves 0.575-s inference with a 0.829 Dice coefficient under optimal conditions. It maintains robust performance down to −5-dB SNR, with monotonic improvements in mean squared error (MSE), mean absolute error (MAE), and Dice as SNR increases. The experimental results indicate that the approach facilitates rapid and accurate retrieval of information on concealed dangerous objects in resource-limited settings.


  • Automatic Localization of Internal Crack in Concrete Structures Based on Hyperbolic Trajectory Extraction From 3-D GPR Data

    by Tian Lan, Zexi Wang, Shiwen Sheng, Jiaxin Yang, Junbo Gong, Leng Liao

    Concrete structures constitute a critical component of infrastructure, including buildings, bridges, and tunnels. However, the initiation and propagation of internal cracks in concrete structures pose substantial challenges to public safety. Ground penetrating radar (GPR) serves as one of the primary tools for detecting internal cracks. However, the endpoint echoes of cracks acquired by GPR are often susceptible to clutter interference. Researchers typically cannot extract the precise locations of echoes from low signal-to-noise ratio (SNR) images, and conventional clutter suppression methods fail to achieve satisfactory results. Consequently, the inversion of 3-D crack information proves challenging. To address the issue of echo extraction from low-SNR images, this article proposes a method for identifying potential target regions from a 3-D slice perspective. The approach leverages the more comprehensive crack information available in 3-D slices, ensuring that some slices with acceptable SNR levels are included. By selecting a hyperbolic crack region from a high-SNR slice as a template, the method matches potential target regions in other slices with poor SNR. Subsequently, the hyperbolic contours of these regions are extracted, enabling the quantification of 3-D crack information and the reconstruction of crack locations. This method has been successfully validated through both simulation and real experiments. In simulations, for vertical cracks at the depth of 5 and 7 cm, the average errors in vertex horizontal position and relative burial depth were 1.14 and 1.49 cm, respectively. In real experiments, for vertical cracks at the depth of 5 and 7 cm, the average errors of horizontal position and relative buried depth were 1.23 and 0.44 cm, respectively. These results confirm that the proposed algorithm accurately determines the position and morphology of internal cracks, thereby providing robust technical support for the health monitoring of concrete structures.


  • Multifrequency Fusion Imaging Algorithm for Nonuniform Wall Scenarios

    by Xiaolu Zeng, Yuxin Miao, Zixiang Yin, Junbo Gong, Xiaopeng Yang

    Through-the-Wall Radar leverages the excellent penetration capabilities of low-frequency electromagnetic waves to achieve non-invasive and non-destructive 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 multi-frequency fusion imaging algorithm which can handle urban nonuniform walls is proposed in this paper. Firstly, a segmented wall compensation algorithm is proposed for nonuniform wall scenarios to address poor focusing effects. Secondly, this paper proposes a multi-frequency 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.

  • Open Access

    Diameter Inversion for Dense Rebars Based on Dual‐Polarized GPR

    by Shuangying Sun, Junbo Gong, Yuhan Li, Kunao Li, Tian Lan

    Reinforced concrete is an essential material in building engineering. Estimating rebar diameter is crucial for evaluating structural load‐bearing capacity and overall stability, which is of great significance for the structural health monitoring of buildings. Ground‐penetrating radar (GPR), as a nondestructive testing technique, can be used to assess rebar diameter. However, in practical engineering, rebars in concrete are often densely distributed, leading to overlapping electromagnetic echoes from multiple targets. Existing diameter inversion methods based on GPR are primarily designed for isolated single rebar and are not well‐suited for diameter estimation in multitarget scenarios. To tackle this problem, this paper presents a fast multiecho separation method to improve the accuracy of diameter inversion in multirebar scenarios. First, a multirebar scattering model for rebars is established. Subsequently, through improved deconvolution and the orthogonal matching pursuit algorithm, individual rebar echoes within the overlapping echoes are rapidly resolved. Finally, simulation and practical experiments have verified that this method can effectively separate individual rebar echoes from multiple rebar echoes. By integrating the diameter inversion algorithms based on the wideband power ratio and natural resonance frequency, respectively, the accuracy of diameter inversion can be significantly enhanced. The relative error in the inverted diameter has decreased by approximately 23%–58%, and the stability of the diameter estimation algorithm has been effectively improved. Compared to the echo separation method using original deconvolution and orthogonal matching pursuit, the enhanced multiecho separation method reduces the computation time by about 80%.

  • Underground Pipeline and Void Recognition in GPR Data: A Nonlearning Method Based on Slope Domain Transformation and Sparse Encoding

    by Tian Lan, Hongchang Chen, Junbo Gong, Chaoyi Huang, Xiaopeng Yang

    Pipeline and void recognition are two key tasks in the underground monitoring of urban roads. Ground penetrating radar (GPR), as an effective geophysical method, plays an important role in this area. With the increasing amount of GPR data, automatic recognition has become a research hotspot. However, existing automated recognition methods still suffer from low accuracy or high dependence on datasets. In this article, a nonlearning method for both pipeline and void recognition is proposed. In this method, the B-scan is preprocessed by removing the direct coupled wave and multiple echoes. Then, the binarized image is converted to sparse image using nonzero interval sparse coding (NISE). Next, the slope distribution of sparse images is extracted using column offset coding (COE). And clustering is carried out according to the corresponding relationship between the image slope and its original position. Finally, the decision is made according to the slope distribution characteristics of each cluster. The proposed method was tested on both simulated and field data. Experimental results show that the method not only has the advantage of being training-free but also exhibits excellent recognition accuracy.


  • A Diffusion Model Based Super-Resolution Imaging Method for Through-Wall Sensing

    by Xiaolu Zeng, Zihan Chen, Xiaopeng Yang, Jiancheng Liao, Junbo Gong

    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 paper 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 multi-source data including the 3D radar images and 2D 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.