•Project presents an application of convolutional neural networks (CNN) in cracks.•Different CNNs are established by the processes of structure design, training and testing.•The crack feature points ...are extracted by feature extraction CNN to establish 3D model.•CNN is able to recognize concealed cracks from other damage in GPR images with zero error.•CNNs could be accurately used for the recognition, location of concealed crack of asphalt pavement.
Concealed cracks in asphalt pavement are the cracks that originate below the surface of the pavement. These cracks are a major contributing factor to pavement damage, in addition to being a major contributing factor to the formation of reflection cracks. The detection of a concealed crack is considered challenging because the location of the crack is, by definition, difficult to find. Therefore, the research on the utilization of ground penetrating radar (GPR) to locate concealed cracks has gained significant interest in recent years. However, the manually processed GPR image used for the recognition, location, and measurement of concealed cracks is inefficient and inaccurate. This project presents an application of convolutional neural networks (CNNs) to GPR images that automatically recognizes, locates, measures, and produces a 3D reconstruction of concealed cracks. In this project, three different CNNs (recognition, location, and feature extraction) were established to accomplish the aforementioned tasks automatically. Each CNN is developed through processes of structural design, training, and testing. The recognition CNN was designed to distinguish concealed cracks from other types of damage in a GPR image, the location CNN determined the location and length measurement of concealed crack images based on the results provided by the recognition CNN, and crack feature points were extracted by the feature extraction CNN to establish the 3D reconstruction models of the concealed cracks. The 3D reconstruction models were then used to calculate crack volume and predict the growth tendency of cracks. The results indicated that the recognition CNN is able to distinguish concealed cracks from other types of damages in 6482 GPR images with zero errors. In addition, the length recognition results calculated from the location CNN possess a 0.2543cm mean squared error, a 0.978cm maximum length error, and a 0.504cm average error in the test samples. Meanwhile, the feature extraction CNN is able to provide feature points for a 3D reconstruction model. The results of this study suggest that the CNNs could be accurately used for the recognition, location, and 3D reconstruction of concealed cracks in asphalt pavement in real-world applications.
Soil water content (SWC) is significant for understanding and evaluating the conditions of soils and plants. Since traditional methods such as time domain reflectometry (TDR) and neutron probes have ...significant drawbacks such as limitations in spatial resolution, detection depth, efficiency, and non-destruction, ground penetrating radar (GPR) has become a potential method in SWC estimation. Many features extracted from GPR data in the time and frequency domain have been proven to be sensitive to the SWC and can further achieve the estimation of it. However, the methods based on these features are easy to be interfered with by noise and the heterogeneity in soils. This article aims to solve this problem by including more features and integrating these features for a joint estimation. Firstly, we study the relationships between SWC and seven features extracted from GPR data. Consequently, we propose to include new features, i.e. the loss tangent feature and the time-frequency features, in the SWC inversion. Secondly, we achieve the multi-feature ensemble learning based on the Adaboost R. method, which largely enhances the accuracy of SWC inversions compared to the single-feature estimations. During the numerical test, we establish the stochastic medium to model the heterogeneity in the real soil. The test verifies the effectiveness and the robustness of the proposed method. Finally, a field experiment is performed on the transition zone of no-tillage and deep-ploughing croplands. A 2-D SWC map is obtained which distinctly presents the SWC difference between the two regions. Our study provides a new approach to improve the accuracy of SWC estimation using GPR.
•A multi-feature ensemble learning method is proposed to inverse soil water content.•Loss tangent is used for the inversion of soil water content.•A two-dimensional soil water content map of a cropland is obtained.•Soil water content in no-tillage and deep-ploughing zones are strongly different.
In this paper, a deep learning-based methodology is proposed to estimate depth and radius for underground pipeline using Ground penetrating radar(GPR) images. The proposed methodology conducts the ...depth and radius joint estimation based on GPR B-scan image preprocessing. During the preprocessing, the autoencoder structure network is designed to binarize the GPR image. Besides, redundant detection boxes of hyperbola detection are also eliminated. For pipeline depth estimation, the parabola model is used instead of the hyperbola model to obtain the symmetry axis of the pipeline GRP image more efficiently. The consecutive a-scan data at extracted symmetry axis and its vicinity are spliced and sent to the Long Short-Term Memory Network for depth estimation. For pipeline radius estimation, the image features and depth information are concatenated. A pre-trained convolutional neural network is designed to extract the GPR image features. Meanwhile, a 2d-Mapping network is proposed to encode the 1-dimensional depth information into 2-dimensional feature matrix. Improving the dimension of depth information rather than flattening the GPR image features (dimension reduction) can preserve the spatial information of image features. Besides, the depth information can be concatenated with the image features more flexibly. In all, the proposed methodology can accurately estimate the depth and radius of underground pipeline. To verify the effectiveness of the proposed methodology, the gprMax is used to generate the pipeline GPR images in inhomogeneous soil and create the dataset. The experiments include detailed comparison and analysis, illustrating that the proposed methodology is feasible and effective.
To better reconstruct underground targets based on ground-penetrating radar (GPR) data, this paper proposes a joint physics and data driven full-waveform inversion (PDD-FWI) scheme. This scheme ...combines a physics-based non-iterative approach and a data-driven deep neural network (DNN) to reconstruct target location, shape and permittivity accurately. Firstly, the normalized range migration algorithm (RMA) is introduced to extract the target contour and location information, which not only improves the robustness of the proposed scheme, but also ensures adaptability to different GPR equipment. Then, the GPR dielectric target reconstruction network (GPRDtrNet) is developed based on the improved U-net structure, including reducing network layers and adding multi-scale additive spatial attention gates and skip-connection structures. Compared with previous DNN-based reconstruction methods, GPRDtrNet has the advantages of small data requirement, high accuracy, strong generalization and noise tolerance. Finally, the simulated and real dataset containing kinds of targets is constructed to train and test GPRDtrNet. The results show that the proposed method can reconstruct underground dielectric targets accurately with high robustness and noise tolerance.
In semi‐arid central Asia, relief has a strong impact on the distribution of vegetation and discontinuous permafrost. Our aim was to analyse causal chains and inter‐relationships that control the ...spatial patterns of forest and permafrost in the forest‐steppe of the northern Khangai Mountains in Mongolia. For this purpose, we conducted soil‐profile descriptions, ground‐penetrating radar sounding, and vegetation mapping to gain information about forest and permafrost distribution. We integrated remote‐sensing analysis and field‐mapping data, including soil properties, vegetation cover, forest fires and anthropogenic forest use. We developed and applied a technique for spatial delineation of permafrost distribution, based on the parameters Topographic Wetness Index (TWI), incoming solar radiation and Normalized Difference Vegetation Index (NDVI). Key outcomes of this study are that the occurrence of discontinuous permafrost within 1 m depth is limited to forest stands larger than 100 ha on north‐facing slopes. Dense ground vegetation supports permafrost, whereas sandy soil texture leads to greater depth of the permafrost table. As the seasonal ice in the active layer progressively melts down during summer, meltwater interflow above the permafrost table provides additional soil moisture downslope. This process is reflected in enhanced vitality of the steppe vegetation on toe slopes below forests with permafrost. This effect can in turn be used to indirectly detect permafrost in forest stands by remote sensing. Permafrost mostly disappears after forest fires and other severe disturbances, but it may re‐establish during forest regrowth. However, climate warming is presently leading to a loss of permafrost regeneration potential after disturbance, and to a shift from climate‐induced and ecosystem‐driven permafrost to entirely ecosystem‐protected permafrost. These trends will result in a further decrease of permafrost area after forest disturbance.
In the forest‐steppe ecotone of Mongolia, meltwater from seasonal ice above permafrost substantially increases the vitality of forest during summer droughts. Vegetation disturbances by forest fire, tree cutting and wood pasture lead to degradation of permafrost. The ecological capacity of permafrost re‐establishment during regrowth of dense forest decreases under actual climate warming.
•An automatic recognition method to detect tunnel lining elements using GPR image.•Deep convolutional networks are used for target recognition.•The FDTD and DCGAN approaches augment the training ...dataset effectively.•High recognition accuracy is achieved on the real GPR dataset.
Tunnel lining inspection using ground penetrating radar (GPR) is a routine procedure to ensure construction quality. Yet, the interpretation of GPR data relies heavily on manual experience that may lead to low efficiency and recognition error when a large volume of data is involved. We introduced a deep learning-based automatic recognition method to identify tunnel lining elements, including steel ribs, voids, and initial linings from GPR images. Based on the mask region-based convolutional neural network (Mask R-CNN), this approach uses the 101-layer deep residual network (ResNet101) with the feature pyramid network (FPN) to extract features, the region proposal network (RPN) to generate candidate regions, a group of fully connected layers to detect the presence and locations of steel ribs and voids, and a fully convolutional network (FCN) to segment the area of the initial lining. To improve the recognition performance of the network, the finite-difference time-domain (FDTD) method and deep convolutional generative adversarial network (DCGAN) are employed to create synthetic GPR images for data augmentation. The test results on a synthetic example show that the mean absolute errors for steel rib, void, and initial lining thickness recognition are 1.2, 2.2, and 4.2mm, respectively, demonstrating the feasibility of the recognition network. In a field GPR survey experiment, the recognition accuracies achieved 96.02%, 91.17%, and 95.45% for the three targets. With the optimal proportions of synthetic images added to the training dataset, the accuracies were further improved to 98.86%, 94.53%, and 99.27%, respectively.
Ground penetrating radar (GPR) has been widely used for detection and localization of reinforced steel bar (rebar) in concrete in a non-destructive way. However, manual interpretation of a large ...number of GPR images is time-consuming, and the results highly depend on practitioner experience and the available priori information. This paper proposes an automatic detection and localization method using deep learning and migration. Firstly, a Single Shot Multibox Detector (SSD) model is established to identify regions of interest containing hyperbolas in a GPR image. This deep learning model is trained using a real GPR dataset, which contains 13,026 rebar targets in 3992 images, collected on residential buildings under construction. Secondly, each target region is migrated and transformed into a binary image to locate the rebar. After the binarization, the apex of the focused cluster is obtained and used to estimate both the horizontal position and the depth of the rebar. The testing results show that the detection accuracy of the proposed artificial intelligence method is 90.9%. The computation time needed for processing a GPR image with a size of 300 × 300 pixels is only 0.47 s. The depth estimation error in a laboratory experiment is <1.5 mm (5%), and the lateral position error is <0.7 cm. Therefore, it is concluded that the proposed method can automatically detect the rebar from GPR images in real time when a handheld GPR system is operated at a walking speed and the depth estimation accuracy is acceptable in practice.
•An automatic algorithm is proposed for localization of rebar in concrete using GPR.•A deep learning algorithm is proposed for automatic detection of rebar hyperbolas.•The SSD model is trained and tested using a real GPR dataset.•The detection speed and accuracy are evaluated.•The rebar depth is estimated with a high accuracy by migration of rebar hyperbola.
Moisture damage is one of the major defects in asphalt pavement, and will evolve into potholes in a short time which will affect traffic safety. Ground Penetrating Radar (GPR) is an effective ...non-destructive testing (NDT) method for detecting moisture damage but its data explanation replies on human experience and subjects to labor intensive. To address this issue, an automatic detection method based on extreme gradient boosting (XGBoost) combined with Bayesian hyper-parameter optimization (BHPO) was proposed to detect moisture damage area from GPR traces. High frequency GPR antenna with 2.3 GHz was used to detect the moisture damage area from simulation, laboratory and field tests, and moisture damage dataset with 7960 traces was collected. Thirty time-frequency parameters were extracted from each GPR trace, normalized to unify the three data source, and then optimized into 12 sensitive parameters by feature importance method. These 12 parameters were used to build the recognition model with XGBoost, and the model tuning parameters were optimized by BHPO. To obtain optimization model, random forest (RF) and artificial neural network (ANN) were also trained with BHPO, and compared with XGBoost model. The results indicate that performance of XGBoost model with BHPO achieves the highest accuracy and lowest time cost both in moisture damage and normal trace classification, the accuracies for moisture damage are XGBoost (96.9%) > ANN (95.6%) > RF (95.4%), respectively, and normal are XGBoost (96.5%) > RF (96.1%) and ANN (96.0%), respectively. On this basis, field tests were conducted by core samples, which verified the correct result of XGBoost model. Our method provides a swift and accurate method to locate subsurface targets directly from GPR signals.
•30 time-frequency features were extracted for representing moisture damage.•12 sensitive features were optimized by feature importance analysis.•XGBoost with BHPO was adopted to locate moisture damage from GPR signal.•Normalizing method was used to combine features from different data source.•BHPO-XGBoost model reaches high accuracy with 96.9%.
•Building a hybrid-polarization GPR system for rebar corrosion detection.•Perpendicular polarization against rebar orientation is more sensitive to rebar corrosion.•Classifications by H-Alpha ...polarization decomposition shift in corrosion process.•Polarimetric GPR shows potential application in evaluation of early-stage corrosion.
As a non-destructive testing method, ground penetrating radar (GPR) has been applied to monitor the corrosion of reinforced bar (rebar) in concrete. However, a traditional GPR system employs a pair of antennas, and can only measure single-polarization radar signals. In this paper, a hybrid-polarization GPR system is proposed for detection and evaluation of rebar corrosion. It can record rebar reflections in two orthogonal polarimetric channels, from which a full-polarimetric scattering matrix can be estimated through polarimetric decomposition and calibration. An accelerated corrosion experiment was conducted and the results prove that radar signal in the polarization channel perpendicular to the rebar orientation is more sensitive to the corrosion, compared with the commonly-used parallel polarization channel. Analysis of the scattering mechanism through H-Alpha decomposition shows that the corroded rebar shifts from the low-entropy dipole scattering to the low-entropy surface scattering. By filling the migrated GPR image with the color of the H-Alpha scattering classification, it is able to visually evaluate the early-stage corrosion of rebar in concrete.