Interpreting Ground Penetrating Radar Data Research Articles

Automatic segmentation of radar data from the Chang’E-4 mission using unsupervised machine learning: A data-driven interpretation approach

Chang’E-3, Chang’E-4 and Chang’E-5 were the first planetary missions with in-situ ground-penetrating radar (GPR) in their scientific payloads. Apart from the Chang’E missions, in-situ GPR has also been used in the Martian missions Tianwen-1 and Perseverance, establishing GPR as a mainstream and pivotal tool in the new era of planetary exploration. Despite its widespread use in planetary science, interpreting GPR data from the Chang’E-4 mission is still challenging, with varying and ambiguous interpretations offered by different researchers. This is primarily due to the complexity and heterogeneity of the Lunar subsurface, which results to a difficult to interpret radagram with low signal-to-clutter ratio. Clutter masks potential targets and makes interpretation a laborious and highly subjective process. To tackle this, we interpret the GPR data from Chang’E-4 mission using a novel processing paradigm based on un-supervised machine learning. A set of statistical attributes is initially calculated locally for different sliding windows in the investigated radagram. Subsequently, these statistical features are clustered into groups using un-supervised machine learning. The segmented radagram highlights boundaries between different formations, and reveals hidden structures previously unseen using typical processing approaches. The proposed framework is fully automatic and objective, facilitating processing, and increasing the overall reliability of the final interpretation.

Review of Ground Penetrating Radar Applications for Bridge Infrastructures

Infrastructure bridges play a crucial role in fostering economic and social development. However, the adverse effects of natural hazard and weather degradation, coupled with escalating rates of traffic, pose a significant threat. The resultant strain on the structure can lead to undue stress, elevating the risk of a critical asset failure. Hence, non-destructive testing (NDT) has become indispensable in the surveillance of bridge infrastructure. Its primary objectives include ensuring safety, optimizing structural integrity, minimizing repair costs, and extending the lifespan of bridges. NDT techniques can be applied to both existing and newly constructed bridge structures. However, it is crucial to recognize that each NDT method comes with its own set of advantages and limitations tailored to specific tasks. No single method can provide an effective and unequivocal diagnosis on its own. Among the various NDT methods, Ground Penetrating Radar (GPR) has emerged as one of the most widely employed techniques for monitoring bridges. In fact, recent technical regulations now mandate the use of GPR for bridge monitoring and characterization, underscoring its significance in ensuring the structural health and longevity of these critical infrastructures. Ground Penetrating Radar (GPR) stands out as one of the most highly recommended non-destructive methods, offering an efficient and timely assessment of the structural conditions of infrastructure. Recognizing the pivotal role of non-destructive testing (NDT) in this context, this paper aims to elucidate recent scientific endeavors related to the application of GPR in bridge engineering structures. The exploration will commence with a focus on studies conducted both at the model level within laboratory settings and on real cases. Subsequently, the discussion will extend to encompass the characterization and monitoring of the bridge’s main elements: slab, beam, and pillar. By delving into these scientific experiences, this paper intends to provide valuable insights into the efficacy and applicability of GPR in assessing and ensuring the structural integrity of bridges. This paper provides a concise survey of the existing literature on the application of Ground Penetrating Radar (GPR) in the assessment of bridges and viaducts constructed with masonry and reinforced concrete, taking into account papers of journal articles and proceedings available on open databases. Various approaches employed in both laboratory and field settings will be explored and juxtaposed. Additionally, this paper delves into discussions on novel processing and visualization approaches, shedding light on advancements in techniques for interpreting GPR data in the context of bridge and viaduct evaluations.

Deep Learning Inversion of Multi-processed GPR Data to Characterize Subsurface Defects

Ground penetrating radar (GPR) is highly recommended for investing defects in concrete structures, yet interpreting GPR data is hindered by the fact that the images primarily display reflected electromagnetic waves rather than the actual shapes of the defects. In this paper, we introduce a deep learning network that aims to convert GPR B-scan images into permittivity maps of subsurface structures. This network takes three types of data as inputs and outputs the distribution of permittivity. These three types of data are the original data, the data obtained after applying migration to the original data, and the data obtained after processing the original data using a neural network. Each type of data is processed using different methods and contains distinct information, which is collectively used for inverting data. In this study, we validate the effectiveness of the proposed network by the finite-difference time-domain (FDTD) forward modeling data.

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.

Improving GPR Imaging of the Buried Water Utility Infrastructure by Integrating the Multidimensional Nonlinear Data Decomposition Technique into the Edge Detection

Although ground-penetrating radar (GPR) is effective to detect shallow-buried objects, it still needs more effort for the application to investigate a buried water utility infrastructure. Edge detection is a well-known image processing technique that may improve the resolution of GPR images. In this study, we briefly review the theory of edge detection and discuss several popular edge detectors as examples, and then apply an enhanced edge detecting method to GPR data processing. This method integrates the multidimensional ensemble empirical mode decomposition (MDEEMD) algorithm into standard edge detecting filters. MDEEMD is implemented mainly for data reconstruction to increase the signal-to-noise ratio before edge detecting. A quantitative marginal spectrum analysis is employed to support the data reconstruction and facilitate the final data interpretation. The results of the numerical model study followed by a field example suggest that the MDEEMD edge detector is a competent method for processing and interpreting GPR data of a buried hot spring well, which cannot be efficiently handled by conventional techniques. Moreover, the proposed method should be readily considered a vital tool for processing other kinds of buried water utility infrastructures.

Recognition of the Typical Distress in Concrete Pavement Based on GPR and 1D-CNN

Ground-penetrating radar (GPR) signal recognition depends much on manual feature extraction. However, the complexity of radar detection signals leads to conventional intelligent algorithms lacking sufficient flexibility in concrete pavement detection. Focused on these problems, we proposed an adaptive one-dimensional convolution neural network (1D-CNN) algorithm for interpreting GPR data. Firstly, the training dataset and testing dataset were constructed from the detection signals on pavement samples of different types of distress; secondly, the raw signals are were directly inputted into the 1D-CNN model, and the raw signal features of the radar wave are extracted using the adaptive deep learning network; finally, the output used the Soft-Max classifier to provide the classification result of the concrete pavement distress. Through simulation experiments and actual field testing, the results show that the proposed method has high accuracy and excellent generalization performance compared to the conventional method. It also has practical applications.

Imaging shallow fault structures by three-dimensional reverse time migration of ground penetration radar data

Ground penetrating radar (GPR) is an effective technology for investigating shallow underground structures using high-frequency electromagnetic waves. Migration is necessary to interpret GPR data because it resolves the true structure from distorted and blurry time-domain profiles. Wave-equation-based reverse time migration (RTM) can provide higher-resolution and more reliable results compared to ray-based migration methods. We have developed a three-dimensional RTM approach based on the finite-difference time-domain algorithm and applied it to field GPR data collected in the northern segment of the Huangzhuang-Gaoliying fault in suburban Beijing. Migrated images clearly show the subsurface structures, with clear indications of underground reflectors in agreement with geological cross-sections proposed in previous studies. Our results demonstrate that GPR is a convenient, non-destructive, and reliable technique that provides a new comprehensive approach to image subsurface conditions, thus allowing effective monitoring of active faults in urban areas.

Data Interpretation Technology of GPR Survey Based on Variational Mode Decomposition

Data interpretation is the crucial scientific component that influences the inspection accuracy of ground penetrating radar (GPR). Developing algorithms for interpreting GPR data is a research focus of increasing interest. The problem of algorithms for interpreting GPR data is unresolved. To this end, this study proposes a sophisticated algorithm for interpreting GPR data with the aim of improving the inspection resolution. The algorithm is formulated by integrating variational mode decomposition (VMD) and Hilbert–Huang transform techniques. With this method, the intrinsic mode function of the GPR data is first produced using the VMD of the data, followed by obtaining the instantaneous frequency by using the Hilbert–Huang transform to analyze the intrinsic mode functions. The instantaneous frequency data can be decomposed into three frequency attributes, including frequency division section, time-frequency section, and space frequency section, which constitute a platform to gain insight into the nature of the GPR data, such that the inspected media components can be examined. The effectiveness of the proposed method on a synthetic signal from a GPR forward model was studied, with the multi-resolution performance being tested. Inspecting the media of a highroad by analyzing the GPR data, with the abnormal characteristics being designated, validated the applicability of the proposed method.

An Application Example of Machine Learning to Shallow Subsurface Exploration using Ground Penetrating Radar

We applied a supervised machine learning method for the purpose of auto detecting cavities under paved roadway from GPR （Ground Penetrating Radar） images. In cavity detection surveys under paved road way, we typically use vehicle-born GPR systems for the necessity to cover wide survey area without any traffic controls. To detect cavity from massive data acquired by vehicle-borne GPR swiftly, skilled-engineers carefully interpret GPR data considering features based on physics of GPR response. Automatic process based on knowledges of skilled-engineers is required, although automatic anomaly detection has not practically realized cause of un-uniform or complex responses from cavities. We applied machine learning methods to detect cavity anomalies using actual GPR data which were acquired on natural occurred cavities as training data. At the veri?cation with actual survey data, our method was able to detect cavitycaused GPR patterns. This will be helpful for analyzer to narrow down cavity responses, meanwhile there were still too many over detections. We think accumulation of labeled GPR data of cavity will also contribute to improvement of our method.

Corrosiveness mapping of bridge decks using image-based analysis of GPR data

Abstract Ground Penetrating Radar (GPR), a locating instrument by design, has attracted much interest as a non-destructive inspection technique for bridge decks due to its ease of use and ability to detect corrosion. The most commonly used technique to interpret GPR data is numerical analysis. The assessment method relies on the fact that corroded reinforcing bars, which cause signal attenuation, can be easily identified using GPR scans. However, several other unrelated factors can attenuate GPR signals such as reinforcing bar depth, surface anomalies and reinforcing bar spacing. These anomalies can be falsely interpreted as deterioration using numerical analysis. Image-based analysis overcomes these drawbacks through analyzing the entire GPR profile while considering prior knowledge of the structure characteristics, thus determining the state of bridge deck. The main objective of this research is to develop a systematic framework of the image-based analysis. The framework is supported by various GPR profiles depicting several causes of signal attenuation and their analysis with respect to deterioration or rebar corrosion status. Two case studies are presented and analyzed using numerical analysis, image-based analysis, and other Non-Destructive Evaluation (NDE) techniques. The results are compared to confirm the validity of the proposed methodology. Further validation is done using concrete cores. The developed approach is believed to assist transportation agencies in a more informed decision making process through highlighting areas of actual deterioration.

GPR Tomographic Imaging of Concrete Tubes and Steel/Plastic Tanks Buried in IAG/USP Geophysical Test Site, Brazil

In this paper, the Ground Penetrating Radar (GPR) method was used to characterize concrete tubes and steel/plastic tanks buried in IAG/USP test site. The microwave tomography was used to improve the GPR images, aiming to retrieve the geometry of the targets. The numerical modeling studies also were done in order to predict the GPR results of the buried targets and to give more reliability to the results interpretation. The targets were installed in the first shallow geophysical test site of the Brazil located at Institute of Astronomy, Geophysics, and Atmospheric Science (IAG) of the University of Sao Paulo (USP). GPR profiles of 200 MHz (shielded bistatic antennas) were acquired along three lines containing concrete tubes and steel/plastic tanks buried in subsoil. The concrete tubes show a hyperbolic reflector for the top, and the vertical tube also presented a reflection on its bottom. The horizontal steel tanks were characterized by a strong GPR reflection on their top. The empty plastic tank shows a strong reflector for the top with normal polarity. On the other hand, the plastic tank filled with water shows a weaker reflector for its top characterized by the inverted polarity of GPR signal when compared with empty plastic tank. The plastic tank filled with water also went characterized by the strong reflection to its bottom, being a good indicative to interpret GPR data on target in subsoil with some types of fluid inside of tank. The results of polarity difference for the top of tank can be used as guide pattern to identify buried tank empty or filled with water. The application of microwave tomography to the GPR data permitted to determine the position and get a good identification of the edges of the targets studied. The numeric modeling presented a good accordance with real data reducing the ambiguities in interpretation of results. These results can be used as a reference, and they can be extrapolated for areas where there is no subsurface information.

Determining velocities in high frequency electromagnetic prospecting by phase shift plus interpolation migration

Phase Shift Plus Interpolation (PSPI) Migration is one of the most popular migration methods that is used not only in Seismic Data Processing but also in interpreting high frequency electromagnetic prospecting [Ground Penetrating Radar (GPR) data]. Based on the similarities between the principle of the propagation of electromagnetic wave and the mechanical wave, migration methods could be applied to interpreting GPR data as a particular step to calculate the medium’s velocity, estimate the depth, shape and size of buried objects. Noticeably, there are two kinds of velocities usually used in migration methods: root mean square (RMS) velocity, which is used in F – K, Finite Difference and Kirchhoff Migration, and interval velocity, which is used in PSPI Migration. RMS velocity is the average velocity taken into account by considering the influence of the upper layer’s instantaneous velocity; whereas the interval velocity only reflect the practical velocity of one layer. In this paper, the problem of how to apply PSPI Migration to interpret GPR data will be presented. Some results of model datum and real datum were also examined. Besides, we made a comparison of using RMS velocity and interval velocity, and then explain how these two types of velocity could be combined to receive the best result.

Content-based image retrieval approaches to interpret ground penetrating radar data

This paper presents a new data processing algorithm to interpret ground penetrating radar (GPR) data for quantification of railroad ballast fouling conditions. The algorithm is based on the observation that different fouling levels generate different textures in the GPR images. The algorithm was designed following the content-based image retrieval procedure, which includes two steps: feature extraction and similarity measurement. First, texture feature was extracted using discrete wavelet transform. Second, similarity measurement was performed. Laboratory GPR data were used to evaluate the accuracy of the algorithm. The accuracy was 93%, which demonstrated the effectiveness of the algorithm.

An automatic recognition algorithm for GPR images of RC structure voids

Ground penetrating radar (GPR) is a powerful tool for detecting defects in and behind reinforced concrete (RC) structures. However, the traditional way of interpreting GPR data involves considerable manpower and is time-consuming. The aim of this study is to illustrate a new approach to recognize GPR images of RC structure voids automatically. Firstly, synthetic GPR images are created by FDTD method. As multiple waves caused by steel bars seriously interfere with the target echo signals, it is difficult to identify targets from the forward modeling images. According to the periodicity of multiple waves from steel bars, the predictive deconvolution method is used to suppress those waves and the outcome is preferable. Then, the support vector machine (SVM) algorithm is proposed to automatically recognize voids in GPR images. The automatic identification procedure includes four steps: 1) collecting training data, 2) extracting features from GPR images, 3) building the SVM model and 4) identifying the voids automatically. The results show that the proposed method provides a suitable tool to locate the cover depths and lateral ranges of the voids, and the trained SVM model gives a favorable outcome when noise (no more than 5%) is added to a synthetic GPR image.

Snow accumulation variability derived from radar and firn core data along a 600 km transect in Adelie Land, East Antarctic plateau

Abstract. The mass balance of ice sheets is an intensively studied topic in the context of global change and sea-level rise. However – particularly in Antarctica – obtaining mass balance estimates remains difficult due to various logistical problems. In the framework of the TASTE-IDEA (Trans-Antarctic Scientific Traverses Expeditions – Ice Divide of East Antarctica) program, an International Polar Year project, continuous ground penetrating radar (GPR) measurements were carried out during a traverse in Adelie Land (East Antarctica) during the 2008–2009 austral summer between the Italian–French Dome C (DC) polar plateau site and French Dumont D'Urville (DdU) coastal station. The aim of this study was to process and interpret GPR data in terms of snow accumulation, to analyse its spatial and temporal variability and compare it with historical data and modelling. The focus was on the last 300 yr, from the pre-industrial period to recent times. Beta-radioactivity counting and gamma spectrometry were applied to cores at the LGGE laboratory, providing a depth–age calibration for radar measurements. Over the 600 km of usable GPR data, depth and snow accumulation were determined with the help of three distinct layers visible on the radargrams (≈ 1730, 1799 and 1941 AD). Preliminary results reveal a gradual increase in accumulation towards the coast (from ≈ 3 cm w.e. a−1 at Dome C to ≈ 17 cm w.e. a−1 at the end of the transect) and previously undocumented undulating structures between 300 and 600 km from DC. Results agree fairly well with data from previous studies and modelling. Drawing final conclusions on temporal variations is difficult because of the margin of error introduced by density estimation. This study should have various applications, including model validation.

Measuring layer thicknesses with GPR – Theory to practice

Ground penetrating radar (GPR) technology has been used to assess pavement performance and structure for the past 30 years in a variety of ways. Yet after all this time, the main issue remains: How well does GPR work and under what conditions? Results show that GPR works well for some situations but not as well for others. It is not currently used on a routine basis by the Departments of Transportation in the US mainly because of difficulties encountered while interpreting GPR data. These difficulties are generally attributed to the fact that the GPR reflected signals that are collected depend largely on the a priori unknown dielectric properties of the structural materials. Additional difficulties arise from the fact that physically GPR cannot detect layers unless they have sufficiently dissimilar dielectric constants. In practice, GPR has been used primarily for pavement layer thickness estimation and moisture accumulation localization within the pavement layers. To improve GPR prediction capabilities, different data processing techniques have been developed that use the GPR reflected signal to estimate the dielectric properties of surveyed structures, thus determining their thicknesses. Other signal processing techniques have also been used successfully to enhance the quality of the GPR signal in order to increase the accuracy of the data interpretation results.

Synthetic and field examples of ground‐penetrating radar (GPR) profile improvement using two‐phase detection techniques

Ground-penetrating radar (GPR) has been used extensively in a wide range of subsurface imaging applications, including underground utilities, hazardous waste and liquid contaminants, buried archeologic sites, geological interfaces (fractures, bedding planes), and pavement–concrete structure assessment (Hugenschmidt, 2000; Olhoeft and Smith, 2000). In the latter case, industrial applications such as damage assessment before road rehabilitation require processing and analysis of data sets that may be orders of magnitude larger than normal GPR ones. In such conditions, the reliable identification of radar echoes in a large range of S/N ratio values can be exploited to apply automatic tracking algorithms for a cost-effective interpretation. A number of studies deal with applying GPR to pavement assessment (Hugenschmidt et al., 1998; Saarenketo and Scullion, 2000), analyzing water content in the pavement or in the subgrade (Benedetto and Benedetto, 2002; Emilsson et al., 2002), and implementing or exploiting automatic tracking algorithms to interpret GPR data sets from pavement surveys Spagnolini and Rampa, 1999). Anomalous lateral variations in water content of the pavement and subgrade are recognized as a major source of pavement damage. At the same time, they affect the radar response by causing an overall decrement of the S/N ratio, most probably because of the defects (primarily delamination and cracks) produced by the anomalous saturation. No examples exist in the available literature of dedicated processing techniques to improve the detection of weak transients in low S/N ratio conditions and to enhance the performance of automatic tracking methods. In this work we exploit wavelet transform-based instantaneous parameter analysis for this purpose and compare the results with those obtained by Hilbert transforms on synthetic and real GPR data.

The GPR refraction method

Refracted phases have been a recognized feature in wide‐angle common midpoint (CMP) and common shotpoint (CSP) ground‐penetrating radar (GPR) surveys for many years, but historically their use has been limited and overshadowed by conventional reflection profiling techniques. In truth, the GPR refraction method holds great promise, as we show using four field examples from three sites in the northeastern United States. Typical refractors are shallow bedrock or coarse‐grained unconsolidated sediments (where the dielectric constant K ≈ 6–7) beneath partially saturated fine‐grained sand or silt (K ≈ 15–16). Under appropriate conditions, refracted phases can be detected readily at transmitter–receiver (Tx–Rx) offsets of >15 m and from depths of >2 m. Adapting procedures from refraction seismology, we interpret GPR data from reversed CSP soundings using dipping planar interface models and delay time analyses. Results from the two approaches are quite consistent, further corroborated at one of the sites by a GPR reflection profile and directly verified by a series of hammer‐driven probes. In addition to refractions from higher velocity layers in the subsurface, the air refracted phase (the ubiquitous phase refracted from the air–earth interface) can provide important constraints on interpreting the subsurface—particularly useful when attempting to resolve a shallow interface (depth ≤ 0.5 m) whose GPR signature might be masked by interacting phases at early traveltime and small transmitter–receiver offsets. Finally, a cautionary note for interpreting conventional reflection profile data when delineating a higher velocity interface at a depth that might be on the order of the antenna separation or less. Since the refracted phase arrives earlier than the reflected phase for Tx–Rx offsets beyond the critical distance, in certain cases the first break on a reflection radargram may actually be a refraction. CMP or CSP soundings in support of conventional GPR profiles can resolve such concerns.

Coincident Antenna Three‐Dimensional GPR

Coincident antenna three‐dimensional ground penetrating radar (GPR) consists of making measurements over an area on the surface using transmitting and receiving antennas that are located very close together. Measurements may be made using antennas that are towed behind a vehicle, or in a fixed‐station (fixed‐trace) mode of operation. In either case, the accurate location of each trace is critical to producing accurate 3D displays. Obtaining a good three‐dimensional display is a critical part of interpreting GPR data. Targets of interest are generally easier to identify and isolate on three‐dimensional data sets than on conventional two‐dimensional profile lines. Simplifying the image by eliminating the noise and clutter is the most important factor for optimizing the interpretation. Image simplification may be achieved by: 1) carefully assigning the amplitude‐color ranges, 2) displaying only one polarity of the GPR signal, 3) using a limited number of colors, 4) decreasing the size of the data set that is...