Ground Penetrating Radar B-scan Research Articles

Training data augmentations for improving hyperbola recognition in ground penetrating radar B-scan image for tree roots detection

Improving the detection accuracy of hyperbola in B-scan images has been a considerable challenge for ground penetrating radar (GPR) to detect tree roots. In this paper, a method for data enhancement and target detection, both based on deep learning was proposed to identify hyperbolas in GPR B-scan images. First, the authors used a cyclic consistent adversarial network (CycleGAN) to augment the original data. In this procedure, the hyperbolic features of the images were preserved and created a wider variety of training samples. Then, the authors could apply the enhanced dataset to the YOLOv5 detection model to evaluate the effectiveness of their method. Meanwhile, the detection effects of Yolov3, Yolov5, Faster R-CNN, and CenterNet detection models on the enhanced dataset were compared. The results showed that applying the enhanced dataset to the Yolov5 detection model exhibited better detection accuracy compared to other combinations of datasets and detection models. The authors demonstrate that the proposed method increases data diversity and the number of samples, improving the precision and recall of hyperbolic curves. These results provide a new method for tree root localization with important effects.

Simultaneous tunnel defects and lining thickness identification based on multi-tasks deep neural network from ground penetrating radar images

The overall assessment of tunnel lining, including shapes, categories, and depths of tunnel internal defects as well as the thickness of tunnel linings is vital to the safe operation of tunnels. We proposed a method comprising a multi-task deep neural network and curve fitting post-processing operation for simultaneously identifying the shapes, categories, and depths of tunnel defects as well as lining thicknesses from ground penetrating radar (GPR) images. The multi-task deep neural network, denoted as M-YOLACT, was designed to identify defects, lining profiles, and hyperbola shapes simultaneously. We introduced a curve-fitting post-processing operation to calculate the dielectric constant automatically based on the hyperbola shapes and evaluated the defect depths and lining thicknesses. The method was validated by numerical simulations, sandbox, and field tests. The method effectively identified the shapes and classes of tunnel defects as well as the thickness profiles from GPR B-Scan images.

MV-GPRNet: Multi-View Subsurface Defect Detection Network for Airport Runway Inspection Based on GPR

The detection and restoration of subsurface defects are essential for ensuring the structural reliability of airport runways. Subsurface inspections can be performed with the aid of a robot equipped with a Ground Penetrating Radar (GPR). However, interpreting GPR data is extremely difficult, as GPR data usually contains severe clutter interference. In addition, many different types of subsurface defects present similar features in B-scan images, making them difficult to distinguish. Consequently, this makes later maintenance work harder as different subsurface defects require different restoration measures. Thus, to automate the inspection process and improve defect identification accuracy, a novel deep learning algorithm, MV-GPRNet, is proposed. Instead of traditionally using GPR B-scan images only, MV-GPRNet utilizes multi-view GPR data to robustly detect regions with defects despite significant interference. It originally fuses the 3D feature map in C-scan data and the 2D feature map in Top-scan data for defect classification and localization. With our runway inspection robot, a large number of real runway data sets from three international airports have been used to extensively test our method. Experimental results indicate that the proposed MV-GPRNet outperforms state-of-the-art (SOTA) approaches. In particular, MV-GPRNet achieves F1 measurements for voids, cracks, subsidences, and pipes at 91%, 69%, 90%, and 100%, respectively.

DMRF-UNet: A Two-Stage Deep Learning Scheme for GPR Data Inversion Under Heterogeneous Soil Conditions

Traditional ground-penetrating radar (GPR) data inversion leverages iterative algorithms that suffer from high computation costs and low accuracy when applied to complex subsurface scenarios. Existing deep learning-based methods focus on the ideal homogeneous subsurface environments and ignore the interference due to clutters and noise in real-world heterogeneous environments. To address these issues, a two-stage deep neural network (DNN), called DMRF-UNet, is proposed to reconstruct the permittivity distributions of subsurface objects from GPR B-scans under heterogeneous soil conditions. In the first stage, a U-shape DNN with first multi-receptive-field convolution (MRF-UNet1) is built to remove the clutters due to inhomogeneity of the heterogeneous soil. Then, the denoised B-scan from MRF-UNet1 is combined with the noisy B-scan to be inputted to the DNN in the second multi-receptive-field convolution (MRF-UNet2). MRF-UNet2 learns the inverse mapping relationship and reconstructs the permittivity distribution of subsurface objects. To avoid information loss, an end-to-end training method combining the loss functions of two stages is introduced. A wide range of subsurface heterogeneous scenarios and B-scans are generated to evaluate the inversion performance. The test results in the numerical experiment and the real measurement show that the proposed network reconstructs the permittivities, shapes, sizes, and locations of subsurface objects with high accuracy. The comparison with existing methods demonstrates the superiority of the proposed methodology for the inversion under heterogeneous soil conditions.

A CNN model for predicting size of buried objects from GPR B-Scans

A convolutional neural networks (CNN) model for predicting size of buried objects from ground penetrating radar (GPR) B-Scans is proposed. As a pre-processing step, Sobel, Laplacian, Scharr, and Canny operators are used for edge detection of the hyperbolic features. The proposed CNN architecture extracts high level signatures in the initial stages of the model and learns additional low-level features when the input data passes through the neural network to finally make an estimation of the required parameter. Artificially generated GPR B-Scans are used to train the model. The proposed method demonstrates good performance in predicting buried object size. Upon comparison, Scharr operator followed by a deep CNN model showed the best performance, having the minimum mean absolute percentage error of 6.74 when tested on new, unseen data.

Underground Object Classification Using Deep 3-D Convolutional Networks and Multiple Mirror Encoding for GPR Data

Ground-penetrating radar (GPR) is an effective tool for underground object detection, but its data interpretation remains a great challenge. In this letter, we propose a novel underground object classification algorithm using deep 3-D convolutional networks (C3D) and multiple mirror encoding (MME) for 3-D GPR data. Although deep learning technique has been applied to interpret the GPR data, most of the existing methods are based on GPR B-scans and have a relatively low accuracy since the reflections from various subsurface targets present similar hyperbolic patterns in B-scans. To improve the classification accuracy, we use 3-D GPR data as training data for C3D to capture the spatio-temporal features between parallel B-scans. Since 3-D GPR data including single object has different sizes in consideration of actual sizes of objects, they are rearranged by the MME method to enhance spatio-temporal features, as well as to satisfy the requirement of the network input. Experimental results demonstrate that the proposed method outperforms the state-of-the-art B-scan-based methods.

A New B-Scan Interpretation Model in Complex Environments

In the scenario where the echo hyperbolas of different targets intersect or the opening structure of the hyperbola is destroyed, the current B-scan image interpretation model has the problems of missing detection and unabling to make full use of the hyperbolic information. The hyperbola needs to be identified manually or by utilizing neural network methods. The recognition accuracy is difficult to guarantee. In order to solve the above problem, a new Ground-Penetrating Radar (GPR) B-scan interpretation model based on hyperbolic trend is proposed in this paper. The model consists of four parts, including preprocessing method, hyperbolic trend clustering algorithm (HTCA), restricted algebraic distance fitting and cluster center algorithm (RADFCC) and Hierarchical Filtered Back Projection (HFBP). Firstly, the GPR B-scan image is converted into a binarized image based on the ratio of the main lobe and side lobe of the echo. Secondly, the points in the binarized image are classified into hyperbolic column-segment clusters based on the hyperbolic trend, and further denoised according to the main lobe width of the ricker wave. Thirdly, by analyzing the characteristics of the GPR in the ground coupling mode, the restricted hyperbola fitting algorithm is used to fit the classified hyperbola column-segment clusters to speed up the convergence of the fit and avoid the result from falling into a local optimum. Fourthly, the target’s location is preliminarily estimated by utilizing the parameters of the hyperbola and the cluster center is established through the location and radius of the target to obtain the accurate estimation of the target and the dielectric constant of the underground. Finally, an imaging algorithm is applied to invert the subsurface image based on the dielectric constant of the subsurface medium. The effectiveness of the proposed model has been demonstrated on both simulated and real datasets.

A Deep Learning-Based GPR Forward Solver for Predicting B-Scans of Subsurface Objects

The forward full-wave modeling of ground-penetrating radar (GPR) facilitates the understanding and interpretation of GPR data. Traditional forward solvers require excessive computational resources, especially when their repetitive executions are needed in signal processing and/or machine learning algorithms for GPR data inversion. To alleviate the computational burden, a deep learning-based 2D GPR forward solver is proposed to predict the GPR B-scans of subsurface objects buried in the heterogeneous soil. The proposed solver is constructed as a bimodal encoder-decoder neural network. Two encoders followed by an adaptive feature fusion module are designed to extract informative features from the subsurface permittivity and conductivity maps. The decoder subsequently constructs the B-scans from the fused feature representations. To enhance the network's generalization capability, transfer learning is employed to fine-tune the network for new scenarios vastly different from those in training set. Numerical results show that the proposed solver achieves a mean relative error of 1.28%. For predicting the B-scan of one subsurface object, the proposed solver requires 12 milliseconds, which is 22,500x less than the time required by a classical physics-based solver.

A Novel Method of Hyperbola Recognition in Ground Penetrating Radar (GPR) B-Scan Image for Tree Roots Detection

Ground penetrating radar (GPR), as a newly nondestructive testing technology (NDT), has been adopted to explore the spatial position and the structure of the tree roots. Due to the complexity of soil distribution and the randomness of the root position in the natural environment, it is difficult to locate the root in the GPR B-Scan image. In this study, a novel method for root detection in the B-scan image by considering both multidirectional features and symmetry of hyperbola was proposed. Firstly, a mixed dataset B-Scan images were employed to train Faster RCNN (Regions with CNN features) to obtain the potential hyperbola region. Then, the peak area and its connected region were filtered from the four directional gradient graphs in the proposed region. The symmetry test was applied to segment the intersecting hyperbolas. Finally, two rounds of coordinate transformation and line detection based on Hough transform were employed for the hyperbola recognition and root radius and position estimation. To validate the effectiveness of this approach for tree root detection, a mixed dataset was made, including synthetic data from gprMax as well as field data collected from 35 ancient tree roots and fresh grapevine controlled experiments. From the results of hyperbola recognition as well as the estimation for the radius and position of the root, our method shows a significant effect in root detection.

GPR B-Scan Image Denoising via Multi-Scale Convolutional Autoencoder with Data Augmentation

Ground-penetrating radar (GPR), as a non-invasive instrument, has been widely used in civil engineering. In GPR B-scan images, there may exist random noise due to the influence of the environment and equipment hardware, which complicates the interpretability of the useful information. Many methods have been proposed to eliminate or suppress the random noise. However, the existing methods have an unsatisfactory denoising effect when the image is severely contaminated by random noise. This paper proposes a multi-scale convolutional autoencoder (MCAE) to denoise GPR data. At the same time, to solve the problem of training dataset insufficiency, we designed the data augmentation strategy, Wasserstein generative adversarial network (WGAN), to increase the training dataset of MCAE. Experimental results conducted on both simulated, generated, and field datasets demonstrated that the proposed scheme has promising performance for image denoising. In terms of three indexes: the peak signal-to-noise ratio (PSNR), the time cost, and the structural similarity index (SSIM), the proposed scheme can achieve better performance of random noise suppression compared with the state-of-the-art competing methods (e.g., CAE, BM3D, WNNM).

Reverse-Time Migration Imaging of Ground-Penetrating Radar in NDT of Reinforced Concrete Structures

The evaluation and inspection of steel bars in reinforced concrete structures are critical for prolonging the service life of buildings. In this regard, ground-penetrating radar (GPR) has been a crucial alternative due to its non-invasiveness and convenience. This paper reports the experimental activities on a test-site area inside a camp in Shanghai, China. To assess the concrete structures of the building, GPR was employed for the detection and localization of rebars in columns, beams, and floors. From the GPR B-scan profiles acquired using a high-frequency antenna, the exact quantity of reinforcements was identified according to the hyperbola responses. Considering the difficulty of inferring the exact position and the scale of the rebars, we applied reverse time migration (RTM) to collapse the hyperbolic response and retrieve the target in a migrated image. To verify the effectiveness of the RTM algorithm, we carried out an experiment on a concrete model with three reinforced bars. We also utilized the RTM algorithm to process the B-scan profiles collected in a column that was later excavated. The imaging results validated the capacity of RTM in localizing and shaping rebars. Then, we employed the RTM algorithm for the GPR B-scan data collected from the other column. Based on the imaging profile, the quantity and positions of the rebars were correctly determined. Moreover, the thickness of the protective layer was evaluated according to the migrated result. These results demonstrate that GPR combined with RTM could provide useful foundation data for structural evaluation.

GPR-RCNN: An Algorithm of Subsurface Defect Detection for Airport Runway Based on GPR

Detection of subsurface defects is important for maintaining runway structural health and reliability. A potential solution is to employ a robot equipped with a Ground Penetrating Radar (GPR) to perform subsurface scanning. To automate the inspection process, we develop a subsurface defect detection algorithm which is a deep learning algorithm that fuses 2D planar features in each panel in GPR B-scans and 3D voxel-wise features in GPR C-scan to robustly detect regions with defects even in the presence of significant noises. Named as GPR-RCNN, we have tested our algorithm with real airport runway data collected from three international airports using our runway inspection robot. The experimental results show that our proposed GPR-RCNN achieves superior results when comparing to state-of-the-art techniques. Specifically, our method achieves F1-measures at 62%, 33%, 81%, and 87% for void, crack, subsidence and pipe, respectively.

Stochastic Processes Approach in GPR Applications

In this article, we present a postprocessing tool to gain powered information for the interpretation of the raw data collected in ground-penetrating radar (GPR) scans by simulations. The method can be extended to measurements raw data and various applications. The proposed approach is based on space ensemble and time ensemble correlation functions applied to GPR B-scan raw data and yields higher detailed information for tracking buried objects. The gprMax simulation results for soil and concrete models confirm the high resolution of the proposed method for sites’ allocation and time stamp of underground objects. The method can be further extended by the introduction of stochastic collocations ensembles in computations.

Roadbed defect detection from ground penetrating radar B-scan data using Faster RCNN

Ground penetrating radar (GPR) is the main technical method to detect roadbed subgrade defect. The recognition of subgrade defect is still mainly based on manual interpretation, which requires high professional knowledge of interpreters, resulting in the demand for automatic detection technology. In this paper, a solution for automatic detection of roadbed defect by implementing Faster RCNN with GPR system is presented. We simulated 30000 roadbed defect GPR B-scan data by simulation software gprMax, labeled them appropriately and automatically. Specifically, Faster RCNN was chosen, as a compromise between accuracy and ease of comparison. Preliminary detection results show that the AP (Average Precision) is 0.8067, proving that our simulation for defect is reasonable and reliable. And the Faster RCNN trained on the simulation dataset without any actual data also has excellent performance on the actual GPR data. Our method of detecting defects automatically with CNN can be easily generalized to other GPR tasks, e.g., detecting pipe, horizon extraction. It’s the biggest open-source GPR B-scan dataset as far as we know. Our simulation dataset and trained model will be made available.

A Gans-Based Deep Learning Framework for Automatic Subsurface Object Recognition From Ground Penetrating Radar Data

Ground penetrating radar (GPR) is a well-known useful tool for subsurface exploration. GPR data can be recorded at a relatively high speed in a continuous way with hyperbolas being artifacts and evidence of disturbances in the soil. Automatic and accurate detection and interpretation of hyperbolas in GPR remains an open challenge. Recently deep learning techniques have achieved remarkable success in image recognition tasks and this has potential for interpretation of GPR data. However, training reliable deep learning models requires massive labeled data, which is challenging. To address the challenges, this work proposes a Generative Adversarial Nets (GANs)-based deep learning framework, which generates new training data to address the scarcity of GPR data, automatically learns features and detects subsurface objects (via hyperbola) through an end-to-end solution. We have evaluated our proposed approach using real GPR B-scan images from rail infrastructure monitoring applications and compared this with the state-of-the-art methods for object detection (i.e. Faster-RCNN, Cascade R-CNN, SSD and YOLO V2). The proposed approach outperforms the existing methods with high accuracy of 97% being the mean Average Precision (mAP). Moreover, the proposed approach also demonstrates the good generalizability through cross-validation on independent datasets.

Deep Learning-Based Rebar Clutters Removal and Defect Echoes Enhancement in GPR Images

The clutters of rebar in the ground penetrating radar (GPR) images may mask the echoes of the inner defects under the rebars, which adversely affects the identification of the inner structural defects in the reinforced concrete (RC). In this study, a deep learning-based method for rebar clutters removal and defect echoes enhancement in GPR B-scan images was proposed. The residual-inception blocks and attention modules were designed in the network as per the characteristics of the task. Validation of the method has been performed at three levels: first using synthetic data, next by synthetic data superimposed with actual concrete background noise, and finally using a sandbox model test in a realistic scenario. The results indicate that the proposed method had the ability to effectively remove rebar clutters and reconstruct complete defect echoes, with performance superior to those of the traditional deep-learning based methods. A RC structural defect identification experiment was performed to verify the effectiveness of the proposed method. The results indicated that after the rebar signals were removed using the method proposed, the accuracy for identifying the defects under the rebars had been improved from 0.208 to 0.850. The proposed method is promising for applying to practical engineering due to its fast processing and better performance.

Underground Cylindrical Objects Detection and Diameter Identification in GPR B-Scans via the CNN-LSTM Framework

Ground penetrating radar (GPR), as a non-invasive instrument, has been widely used in the civil field. The interpretation of GPR data plays a vital role in underground infrastructures to transfer raw data to the interested information, such as diameter. However, the diameter identification of objects in GPR B-scans is a tedious and labor-intensive task, which limits the further application in the field environment. The paper proposes a deep learning-based scheme to solve the issue. First, an adaptive target region detection (ATRD) algorithm is proposed to extract the regions from B-scans that contain hyperbolic signatures. Then, a Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) framework is developed that integrates Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) network to extract hyperbola region features. It transfers the task of diameter identification into a task of hyperbola region classification. Experimental results conducted on both simulated and field datasets demonstrate that the proposed scheme has a promising performance for diameter identification. The CNN-LSTM framework achieves an accuracy of 99.5% on simulated datasets and 92.5% on field datasets.

Clutter Suppression in GPR B-Scan Images Using Robust Autoencoder

Ground-penetrating radar (GPR) is a well-known geophysical electromagnetic method used to detect the underground facilities such as landmines, pipelines, and cavities. In general, the clutter presented in GPR B-scan image obscures the underground objects, thus damaging the performance of the underground object detection algorithm. In this letter, we proposed a new clutter suppression method based on robust autoencoder (RAE). The proposed algorithm decomposes a GPR B-scan image into its low-rank and sparse components. The low-rank component catches the clutter, whereas the sparse component captures the underground object responses. The commonly used clutter removal algorithms, mean subtraction (MS), singular value decomposition (SVD), robust principal component analysis (RPCA), and morphological component analysis (MCA), are compared with the proposed algorithm on both the numerical simulated data and real GPR data. The visual and quantitative results demonstrate the effectiveness of the proposed RAE-based algorithm over the widely used state-of-the-art clutter removal algorithms.

Combined Migrations and Time-Depth Conversions in GPR Prospecting: Application to Reinforced Concrete

In this paper, we propose the combination of different migration results achieved on the same data in order to account for different values of the propagation velocities of the electromagnetic waves within the considered Ground Penetrating Radar (GPR) profile. These different values can be the result of a variable lithological composition or (more probably for short Bscans) the results of different moisture levels, or both. Here, we separately consider the two cases of horizontal or vertical variability of the propagation velocity with a transition zone between two zones with constant propagation velocity. Moreover, we also propose a time-depth conversion accounting for these different values of the propagation velocity along the considered GPR Bscan. The method is applied to real data gathered in the field with regard to a concrete coverage containing liner layers.

NDT assessment of rigid pavement damages with ground penetrating radar: laboratory and field tests

ABSTRACT Cracks in rigid pavements are a frequent and critical problem, which needs to be assessed to avoid deterioration and provide comfort and security. Control of damage requires the evaluation of cracks width and depth. Several laboratory tests were used in order to determine the ability of Ground Penetrating Radar (GPR) in the evaluation of damage under different controlled conditions. Those tests were focused in the analysis of different crack dimensions, changing the material inside the fractures and the orientation of the antenna with respect the crack axis. Results showed that the crack width increases hyperbolae amplitude, being the thinnest crack detected with a 2.3 GHz antenna, 2 mm thick. GPR B-scans were compared to images from computational models. Additionally, a field survey provided images according to the lab tests.