Real Ground Penetrating Radar Data Research Articles

The ground penetrating radar response recognition of hidden dam defects using multi-output convolutional neural network

The Ground Penetrating Radar (GPR) data interpretation of dam defects detection by manual process is a heavy task. Although the convolutional neural network (CNN) is applicable to detect road defects with the help of other auxiliary instruments and techniques, it is still a challenge for the dam defect detection by deep learning. To overcome this problem, a multi-output CNN model for GPR response recognition of dam defects is proposed. The training dataset and test dataset are produced from a large number of GPR responses with different types of dam defects forward modeling by GPRMax3.0 software. The results show that the highest recognition accuracy of the training dataset and test dataset is about 98% and 97%, respectively. To verify the effectiveness, the new proposed method is used to recognize the hidden GPR responses of dam defects in the real GPR data. The results show that the types and locations of dam defects in the real GPR data can be recognized and classified automatically, efficiently and accurately.

3D imaging and temporal evolution recognition of concrete internal defects based on GPR

Periodic inspections and recognition of temporal evolution for concrete internal defects are important for the long-term operation of civil engineering infrastructures. This study proposes a method integrating convolutional neural network (CNN) with maximum intensity projection (MIP) for 3D imaging and temporal evolution recognition of concrete internal defects. The proposed method utilizes MIP to process multiple ground penetrating radar (GPR) B-scans and generates 2D projected GPR B-scans containing spatial information. SegNet coupled with the Lovász softmax loss function is introduced to reconstruct 2D defects from projected GPR B-scans. Subsequently, 3D imaging result is reconstructed from 2D imaging results by 3D reconstruction module. Finally, the defect change extraction module realizes recognition of defect changes based on 3D imaging results at different time. The superiority of proposed method is validated based on both synthetic and real GPR data, which presents better recognition results and less time consumption than existing 3D CNN-based method.

GCNet: Ground Collapse Prediction Based on the Ground-Penetrating Radar and Deep Learning Technique

Manual methods to detect and identify potential ground collapse hazards have a high volume of workload and rely on manual skills and experience strongly, with substantial inconsistencies. This paper proposes an automatic recognition method of potential underground hazards and designs an object detection-based algorithm to predict the hidden danger of ground collapse, called GCNet. The GCNet is a ground collapse hazard detection network, which is developed to detect underground hazards, including cavities, poor soil quality, pipelines, and soil background. The GCNet uses the deep residual network ResNet101 and feature pyramid network (FPN) to extract features and a task coordination network (TCN) to identify the category and location of hidden underground hazards. Further, a feature enhancement method based on the regional binary pattern is proposed to improve the accuracy of the proposed model by expanding the ground-penetrating radar (GPR) data by adding multi-level features to GPR images. The experiments are conducted on a large amount of real GPR data and experiment results show that the proposed automatic recognition method can surpass the existing deep learning-based methods in hidden underground hazard recognition and identification.

Evaluating the optimised YOLO-based defect detection method for subsurface diagnosis with ground penetrating radar

Ground-penetrating radar (GPR) is an efficient and effective non-destructive method for diagnosing urban roads, but its interpretation requires complex analysis. To address this, we proposed an optimized YOLO-based framework for timely identification of road defects using GPR. We used transfer learning and data augmentation to optimize the framework and evaluated their effects. We compared six YOLO versions and found that YOLOv5_s performed best with less weight. The optimized YOLOv5_s-based framework was proposed to identify voids and separation between road layers using more than 100 km of real GPR data. We validated the framework by comparing it with professional visual interpretation and observed that it provided comparable accuracy within seconds. The study benchmarks YOLOv5_s for timely road inspection with GPR.

Efficient Underground Target Detection of Urban Roads in Ground-Penetrating Radar Images Based on Neural Networks

Ground-penetrating radar (GPR) is an important nondestructive testing (NDT) tool for the underground exploration of urban roads. However, due to the large amount of GPR data, traditional manual interpretation is time-consuming and laborious. To address this problem, an efficient underground target detection method for urban roads based on neural networks is proposed in this paper. First, robust principal component analysis (RPCA) is used to suppress the clutter in the B-scan image. Then, three time-domain statistics of each A-scan signal are calculated as its features, and one backpropagation (BP) neural network is adopted to recognize A-scan signals to obtain the horizontal regions of targets. Next, the fusion and deletion (FAD) algorithm is used to further optimize the horizontal regions of targets. Finally, three time-domain statistics of each segmented A-scan signal in the horizontal regions of targets are extracted as the features, and another BP neural network is employed to recognize the segmented A-scan signals to obtain the vertical regions of targets. The proposed method is verified with both simulation and real GPR data. The experimental results show that the proposed method can effectively locate the horizontal ranges and vertical depths of underground targets for urban roads and has higher recognition accuracy and less processing time than the traditional segmentation recognition methods.

Loss Tangent Estimation From GPR Data in Complex Medium Scenarios Using Amplitude and Power Spectrum Centroid Frequency Shift Methods

Electromagnetic losses estimation in complex scenarios lacking subsurface continuous reflectors is challenging. We present here a robust approach based on the centroid frequency shift method and Ricker wavelet as the input signal, to compute intrinsic (absorption) losses. The centroid-frequency shift is calculated using the amplitude and power spectrum of records, leading to equations that relate the centroid frequency variation in time to the loss tangent (intrinsic losses). First, the approaches are evaluated and compared by implementation to synthetic Ground Penetrating Radar (GPR) data generated with the gprMax simulator, starting from an inhomogeneous material with spatially smooth dielectric variations. Then, they are applied to real GPR data collected on a fractured basaltic lava flow. The retrieved loss tangent values are compared to the total attenuation estimated from the signal amplitude decay, to evaluate the contribution due to scattering phenomena. The proposed analysis is particularly suitable for GPR data collected in planetary environments, since dielectric losses could be used to constrain the oxide composition of the material as well as scattering losses to determine the amount and distribution of subsurface inhomogeneities.

An Automated GPR Signal Denoising Scheme Based on Mode Decomposition and Principal Component Analysis

Ground penetrating radar (GPR) signal processing has used mode decomposition methods such as empirical mode decomposition (EMD), variational mode decomposition, and their variations, but these methods are typically constrained by manual mode selection for the extraction of real signals and unwanted noises. In this study, we propose an automatic mode selection scheme for GPR signal denoising, which is demonstrated by the combination of the improved complete ensemble EMD (ICEEMD) and the multiscale principal component analysis (MSPCA). For GPR B-scan data, the intrinsic mode functions (IMFs) are automatically screened using our predefined criteria to remove noise after their spectra and the relationship between signals are first analyzed. Next, the nonlinear and nonstationary signals are adaptively decomposed into many time series by the proposed automated ICEEMD. Each A-scan dataset is then screened and subjected to MSPCA before being reassembled into the final denoised result. The synthetic and real GPR data prove our denoising scheme’s superiority over more established denoising techniques like MSPCA and median filtering.

Using dictionary learning for clutter reduction in GPR B-scan images

ABSTRACT Ground-penetrating radar (GPR) has been widely used to detect subsurface objects. However, the target reflection experiences interference owing to clutter, such as direct coupling, interface reflection, and other undesired reflections. Reducing such clutter is a valuable processing step to improve detection accuracy. This letter proposes a GPR clutter reduction algorithm based on dictionary learning (DL). DL learns an adaptive dictionary that can sparsely represent the GPR B-scan image and enable clutter suppression. The learned dictionary atoms are divided into target atoms and clutter atoms, the target component and the clutter component can be reconstructed using these two types of dictionary atoms, respectively. The performance of the proposed algorithm is validated using both simulation data and real GPR data. The results of the visual evaluation and quantitative analysis demonstrate that the proposed algorithm outperforms the existing clutter reduction algorithms.

Karstified zone interpretation using deep learning algorithms: Convolutional neural networks applications and model interpretability with explainable AI

Ground penetrating radar (GPR) is a non-invasive geophysical method that can be used to assist in mapping karstified zones in outcrop analogous for the characterization and understanding of carbonate oil reservoirs. Supported by GPR data, it is possible to understand the behavior of karstification processes in carbonate rocks and thus expand the knowledge to the reservoir level. In addition, Machine Learning (ML) and Deep Learning (DL) algorithms have allowed the application of computer vision techniques to identify geological structures and facies based on geophysical data, but mostly focused on seismic data, while GPR based applications still need to be explored. In this context, this study seeks to apply deep learning models in synthetic as well as real GPR data and attributes (similarity, energy, instantaneous frequency, instantaneous phase and Hilbert trace/similarity) based on convolutional neural networks using the U-Net architecture capable of assisting in the mapping of karstified zones imaged through GPR surveys. Moreover, explainable artificial intelligence (XAI) techniques using SHapley Additive exPlanation (SHAP) values are applied to improve the interpretability and explainability of the generated models. These techniques were employed in order to assess the rules found by the models, the modeling quality and the presence of biases in the model. The SHAP values show that the energy attribute was the feature that provided more information in both synthetic and real models. Instantaneous frequency and similarity were the second and third most important features in the synthetic model, while the other attributes presented less relevant contributions. Finally, we generated a model capable of aiding in mapping karstified zones using GPR data and attributes, promoting the understanding of complex models and allowing greater cooperation between experts in the geosciences and results generated through deep learning techniques.

Investigating Transfer Learning Performances of Deep Learning Models for Classification of GPR B-Scan Images

Recent advances in deep learning models have made them the state-of-the art method for image classification. Due to this success, they have been applied to many areas, such as satellite image processing, medical image interpretation, video processing, etc. Recently, deep learning models have been utilized for processing Ground Penetrating Radar (GPR) data as well. However, studies general focus on building new Convolutional Neural Network (CNN) models instead of utilizing baseline ones. This paper investigates the usefulness of existing baseline CNN models for classifying GPR B-scan images and aims to determine how well pre-trained models perform. To that end, a real bridge deck GPR data, DECKGPRHv1.0 dataset was used to evaluate the transfer learning performances of various CNN models. Different variants of the models in terms of varying depths and number of parameters were also considered and evaluated in a comparative manner. Although it is an older model, ResNet achieved the best results with 0.998 accuracy. The experimental results showed that there is generally a direct correlation between the simplicity of the model and its success. Overall, it is concluded that near perfect results are possible by just adapting pre-trained models to the problem without fine-tuning.

Loss tangent estimation from ground-penetrating radar data using Ricker wavelet centroid-frequency shift analysis

A novel method to estimate the loss tangent using ground-penetrating radar (GPR) data and assuming a Ricker wavelet as the transmitted radar signal is presented. This method proposes a new relation between centroid frequency and loss tangent, which allows retrieval of the loss tangent of a material through a probabilistic inversion approach. The reliability of the results is carefully evaluated by performing an uncertainty analysis that accounts for the white Gaussian noise affecting the radar data. The method is validated in low- and high-loss scenarios, using synthetic data generated by a well-established finite-difference time-domain model (gprMax). The simulations are performed assuming a dipping sharp reflector separating two materials with different electric properties. The loss tangent values retrieved using the centroid-frequency method are compared, for different levels of noise, with those estimated using the maximum frequency method. The results find that the centroid-frequency method is, in general, more precise when the noise level is high, and it is more accurate in a high-loss environment. Finally, the new approach is applied to real GPR data collected with 500 MHz and 1 GHz antennas on a volcanic ash deposit on Mount Etna (Italy), using the signal reflected from a sloping interface separating two distinct volcanic ash layers.

Declutter-GAN: GPR B-Scan Data Clutter Removal Using Conditional Generative Adversarial Nets

Clutter removal in ground-penetrating radar (GPR) B-scan data has been widely studied in recent years. In this letter, we propose a novel data-driven clutter suppression method in GPR data based on conditional generative adversarial nets (cGANs). The proposed method learns a function that maps the cluttered data to the clutter-free data from the training set. The training set consists of pairs of cluttered data and corresponding clutter-free data. Different from the traditional method that only uses the simulation training set, we simulate the clutter-free data and add the real collected non-target data to the simulated clutter-free data as cluttered data, so that the trained network can generalize well to the real GPR data. The proposed method is compared with the subspace method, sparse representation-based method, and low-rank and sparse matrix decomposition (LRSD) methods on both simulation data and real collected data. The results show that the proposed method has higher performance in terms of computational complexity, clutter suppression results, and applicability than those state-of-the-art methods.

GPRInvNet: Deep Learning-Based Ground-Penetrating Radar Data Inversion for Tunnel Linings

A DNN architecture referred to as GPRInvNet was proposed to tackle the challenges of mapping the ground-penetrating radar (GPR) B-Scan data to complex permittivity maps of subsurface structures. The GPRInvNet consisted of a trace-to-trace encoder and a decoder. It was specially designed to take into account the characteristics of GPR inversion when faced with complex GPR B-Scan data, as well as addressing the spatial alignment issues between time-series B-Scan data and spatial permittivity maps. It displayed the ability to fuse features from several adjacent traces on the B-Scan data to enhance each trace, and then further condense the features of each trace separately. As a result, the sensitive zones on the permittivity maps spatially aligned to the enhanced trace could be reconstructed accurately. The GPRInvNet has been utilized to reconstruct the permittivity map of tunnel linings. A diverse range of dielectric models of tunnel linings containing complex defects has been reconstructed using GPRInvNet. The results have demonstrated that the GPRInvNet is capable of effectively reconstructing complex tunnel lining defects with clear boundaries. Comparative results with existing baseline methods also demonstrated the superiority of the GPRInvNet. For the purpose of generalizing the GPRInvNet to real GPR data, some background noise patches recorded from practical model testing were integrated into the synthetic GPR data to retrain the GPRInvNet. The model testing has been conducted for validation, and experimental results revealed that the GPRInvNet had also achieved satisfactory results with regard to the real data.

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.

A statistics-based descriptor for automatic classification of scatterers in seismic sections

Discontinuities and small structures induce diffractions on seismic or ground-penetrating radar (GPR) acquisitions. Therefore, diffraction images can be used as a tool to access valuable information concerning subsurface scattering features, such as pinch outs, fractures, and edges. Usually, diffraction-imaging methods operate on diffraction events previously detected. Pattern-recognition methods are efficient to detect, image, and characterize diffractions. The use of this kind of approach, though, requires a numerical description of image points on a seismic section or radargram. We have investigated a new descriptor for seismic/GPR data that distinguishes diffractions from reflections. The descriptor consists of a set of statistical measures from diffraction operators sensitive to kinematic and dynamic aspects of an event. We develop experiments in which the proposed descriptor was incorporated into a pattern-recognition routine for diffraction imaging. The obtained method is useful for performing the automatic classification of image points using supervised and unsupervised algorithms, as a complementary step to Kirchhoff imaging. We also develop a new type of filtering, designed to address anomalies on the diffraction operators caused by interfering events. We evaluate the method using synthetic seismic data and real GPR data. Our results indicate that the descriptor correctly discriminates diffractions and shows promising results for low signal-to-noise-ratio situations.

A computer vision based rebar detection chain for automatic processing of concrete bridge deck GPR data

Manual processing of Ground Penetrating Radar (GPR) images is a very time-intensive task. The authors proposed a novel computer vision-based method for automatic detection of rebars in complex GPR images in highly deteriorated concrete bridge decks. The proposed detection model consists of a fine-tuned Histogram of Oriented Gradients feature descriptor, a Multi-Layer Perceptron for classification, and a post processing algorithm for eliminating false detections and labeling rebar in Region of Interest. State-of-art results are obtained on testing the method on real bridge deck GPR data and comparing the results with RADAN software. Overall accuracy of 89.4% is obtained on URIGPRv1.0 dataset, which is introduced in this paper. The proposed method is 54.35% more accurate comparing to the results obtained by RADAN software. The proposed classifier outperformed accuracy of a 3-layer convolutional neural network by 11.9%.

An Optimized Choice of UCPML to Truncate Lattices With Rotated Staggered Grid Scheme for Ground Penetrating Radar Simulation

Efficient and accurate simulation of ground penetrating radar (GPR) in the open region helps immensely in both grasping the features of echoes and facilitating the interpretation of real GPR data. Due to the limitation of the computer model, however, the strong artificial boundary reflections, especially the low-frequency propagating waves encountered at the late stage of simulation greatly affect the simulation accuracy of GPR. This paper presents an innovative optimized unsplit-field convolutional perfectly matched layer (UCPML) based on rotated staggered grid (RSG) scheme to truncate the finite-difference time-domain (FDTD) lattices. Rather than obey the sharp variation based on an ${m}$ th-order polynomial, the optimized approach employs a novel optimized term and an adjustment factor to seek a gentle variation on optimal constitutive coefficients. This guarantees that the determination of optimal constitutive coefficients can be less influenced by the order of polynomial and especially, to improve the absorptive performance on low-frequency propagating waves. The calculating efficiency and accuracy of the RSG-FDTD scheme, as well as the absorbing performance of the optimized UCPML, are verified by two numerical examples. In particular, the analysis of the amplitude-frequency features of low-frequency clutters at steady state of the electromagnetic (EM) field and the corresponding global reflection error in the time–frequency domain is also presented.

A Machine Learning-Based Fast-Forward Solver for Ground Penetrating Radar With Application to Full-Waveform Inversion

The simulation, or forward modeling, of ground penetrating radar (GPR) is becoming a more frequently used approach to facilitate the interpretation of complex real GPR data, and as an essential component of full-waveform inversion (FWI). However, general full-wave 3-D electromagnetic (EM) solvers, such as the ones based on the finite-difference time-domain (FDTD) method, are still computationally demanding for simulating realistic GPR problems. We have developed a novel near-real-time, forward modeling approach for GPR that is based on a machine learning (ML) architecture. The ML framework uses an innovative training method that combines a predictive principal component analysis technique, a detailed model of the GPR transducer, and a large data set of modeled GPR responses from our FDTD simulation software. The ML-based forward solver is parameterized for a specific GPR application, but the framework can be applied to many different classes of GPR problems. To demonstrate the novelty and computational efficiency of our ML-based GPR forward solver, we used it to carry out FWI for a common infrastructure assessment application—determining the location and diameter of reinforcement bars in concrete. We tested our FWI with synthetic and real data and found a good level of accuracy in determining the rebar location, size, and surrounding material properties from both data sets. The combination of the near-real-time computation, which is orders of magnitude less than what is achievable by traditional full-wave 3-D EM solvers, and the accuracy of our ML-based forward model is a significant step toward commercially viable applications of FWI of GPR.

Clutter Elimination and Random-Noise Denoising of GPR Signals Using an SVD Method Based on the Hankel Matrix in the Local Frequency Domain.

Ground-penetrating radar (GPR) is a kind of high-frequency electromagnetic detection technology. It is mainly used to locate targets and interfaces in underground structures. In addition to the effective signals reflected from the subsurface objects or interfaces, the GPR signals in field work also include noise and different clutters, such as antenna-coupled waves, ground clutters, and radio-frequency interference, which have similar wavelet spectral characteristics with the target signals. Clutter and noise seriously interfere with the target’s response signal. The singular value decomposition (SVD) filtering method can select appropriate singular values and characteristic components corresponding to the effective signals for signal reconstruction to filter the GPR data. However, the conventional time-domain SVD method introduces fake signals when eliminating direct waves, and does not have good suppression of random noise around non-horizontal phase axes. Here, an SVD method based on the Hankel matrix in the local frequency domain of GPR data is proposed. Different numerical models and real field GPR data were handled using the proposed method. Based on the power of fake signals introduced via different processes, qualitative and quantitative analyses were carried out. The comparison shows that the newly proposed method could improve efforts to suppress random noise around non-horizontal phase reflection events and weaken the horizontal fake signals introduced by eliminating clutter such as ground waves.

Anisotropic Reverse-Time Migration of Ground-Penetrating Radar Data Collected on the Sand Dunes in the Badain Jaran Desert

Ground-penetrating radar (GPR) profiling is the primary tool to provide detail information of the internal structure and characteristics inside a sand dune in the desert area. However, with the severe elevation change in a short horizontal distance and on the rugged and complex surface topography of the sand dunes, getting clear imaging of the internal structure of sand dunes from the GPR profiling is still a big challenge. The classic imaging technique such as the Kirchhoff migration has been applied to process GPR data on sand dunes, with limited success. The reverse-time migration (RTM) technique is the most advanced imaging technique that can handles GPR data acquired on rugged surface with severe topographic relief to generate subsurface structural images with high fidelity and has been tried to process GPR data on sand dunes. The results are encouraging, and the imaging quality is significantly improved. The finite-difference time-domain (FDTD) method is the major numerical tool for forward and backward continuations of the wave field during the RTM process. There are two aspects for using FDTD in RTM of GPR data on sand dunes still need improvement: the numerical scattering caused by the staircase approximation of the ground surface by using gridding in the Cartesian coordinate, and the negligence of the possible anisotropy of the electromagnetic material properties due to the calcareous cementation bedding inside a sand dune. In this paper, we develop the RTM algorithm based on the staggered grid FDTD that handles the rugged topographic surface by using the curvilinear coordinate, and the possible anisotropic radar wave velocity of the sand dune media. We first demonstrate the equivalency of the nonuniform, isotropic medium and the uniform, anisotropic medium for justifying using the uniform, anisotropic velocity in the RTM by synthetic modeling. Next, we validate our approach of using the synthetic data with the comparison of using the Cartesian coordinate and the curvilinear coordinate in an isotropic medium. The results indicate that the RTM algorithm using the curvilinear coordinate can efficiently eliminate the adverse effect of the staircase approximated boundary of the topography surface. Finally, we processed the real GPR data collected on a sand dune in the Badain Jaran desert by using the curvilinear coordinate and the uniform, anisotropic velocity in FDTD forward and backward wave field continuation. Comparison of the results indicates that the RTM imaging using the boundary-conforming curvilinear coordinate and anisotropic velocity gains more coherent and higher resolution images for the calcareous cementation layers and the water table.