Real Ground Penetrating Radar Data Research Articles

Improving the Autoregressive Modeling Method in Random Noise Suppression of GPR Data Using Undecimated Discrete Wavelet Transform

Geophysics has played a significant and efficient role in studying geological structures over the past decades as the goal of geophysical data acquisition is to investigate underground phenomena with the highest possible level of accuracy. The ground penetrating radar (GPR) method is used as a nondestructive method to reveal shallow structures by beaming electromagnetic waves through the Earth and recording the received reflections, albeit inevitably, along with random noise. Various types of noise affect GPR data, among the most important of which are random noise resulting from arbitrary motions of particles during data acquisition. Random noise which exists always and at all frequencies, along with coherent noise, reduces the quality of GPR data and must be reduced as much as possible. Over the recent years, discrete wavelet transform has proved to be an efficient tool in signal processing, especially in image and signal compressing and noise suppression. It also allows for obtaining an accurate understanding of the signal properties. In this study, we have used the autoregression in both wavelet and f-x domains to suppress random noise in synthetic and real GPR data. Finally, we compare noise suppression in the two domains. Our results reveal that noise suppression is conducted more efficiently in the wavelet domain due to decomposing the signal into separate subbands and exclusively applying the method parameters in autoregression modeling for each subband.

A novel method to remove GPR background noise based on the similarity of non-neighboring regions

Ground penetrating radar (GPR) is a non-destructive technique that has been widely used in many areas of research, such as landmine detection or subsurface anomalies, where it is required to locate targets embedded within a background medium. One of the major challenges in the research of GPR data remains the improvement of the image quality of stone materials by means of detection of true anisotropies since most of the errors are caused by an incorrect interpretation by the users. However, it is complicated due to the interference of the horizontal background noise, e.g., the air-ground interface, that reduces the high-resolution quality of radargrams. Thus, weak or deep anisotropies are often masked by this type of noise. In order to remove the background noise obtained by GPR, this work proposes a novel background removal method assuming that the horizontal noise shows repetitive two-dimensional regions along the movement of the GPR antenna. Specifically, the proposed method, based on the non-local similarity of regions over the distance, computes similarities between different regions of the same depth in order to identify most repetitive regions using a criterion to avoid closer regions. Evaluations are performed using a set of synthetic and real GPR data. Experimental results show that the proposed method obtains promising results compared to the classic background removal techniques and the most recently published background removal methods.

Estimating geometrical parameters of cylindrical targets detected by ground-penetrating radar using template matching algorithm

In the current research, the ground-penetrating radar (GPR) method has been employed to identify physical and geometrical parameters of buried cylindrical structures using the pattern recognition approach. To achieve this goal, the well-established mathematical relationships between geometrical parameters of cylindrical target (radius, burial depth, and horizontal location) and the associated GPR hyperbolic response characteristics are employed using the template matching method. In order to validate the applicability of the template matching method in providing estimates of such parameters, the method is first examined on GPR responses of synthetic models with known geometrical parameters followed by applying on real data using two different similarity criteria including 2-D spatial convolution and normalized cross correlation in the wave number domain. In the first step, the GPR responses of 71 synthetic models encompassing one, two, and three horizontal cylinders were produced using the improved 2-D finite difference in frequency domain. Then, appropriate preprocessing sequences to reduce random noise caused by forward modeling were applied on synthetic data. The proposed algorithm applied on several synthetic model responses could estimate the known geometrical parameters of the buried cylinders with acceptable accuracy (maximum error of 15%). The template matching algorithm was also used to extract geometrical parameters of water and wastewater pipes buried in Imam Hossein Square, Isfahan city, as real GPR data. Depending on environmental conditions and subsurface host formation, the real GPR data normally contain a variety of noises; therefore, a series of appropriate objective preprocessing and processing stages were designed in order to apply on real GPR images before deploying template matching algorithm. The applicability of the template matching algorithm on real data and validity of the estimated parameters were proved based on assessing the accuracy of the estimated geometrical parameters of respective pipes through GPR response versus the measured parameters. The proposed algorithm was designed in such a way that all steps of estimating geometrical parameters of buried cylindrical targets are automatically carried out.

Real-Time Hyperbola Recognition and Fitting in GPR Data

The problem of automatically recognizing and fitting hyperbolae from ground-penetrating radar (GPR) images is addressed, and a novel technique computationally suitable for real-time on-site application is proposed. After preprocessing of the input GPR images, a novel thresholding method is applied to separate the regions of interest from background. A novel column-connection clustering (C3) algorithm is then applied to separate the regions of interest from each other. Subsequently, a machine learnt model is applied to identify hyperbolic signatures from outputs of the C3 algorithm, and a hyperbola is fitted to each such signature with an orthogonal-distance hyperbola fitting algorithm. The novel clustering algorithm C3 is a central component of the proposed system, which enables the identification of hyperbolic signatures and hyperbola fitting. Only two features are used in the machine learning algorithm, which is easy to train using a small set of training data. An orthogonal-distance hyperbola fitting algorithm for “south-opening” hyperbolae is introduced in this work, which is more robust and accurate than algebraic hyperbola fitting algorithms. The proposed method can successfully recognize and fit hyperbolic signatures with intersections with others, hyperbolic signatures with distortions, and incomplete hyperbolic signatures with one leg fully or largely missed. As an additional novel contribution, formulas to compute an initial “south-opening” hyperbola directly from a set of given points are derived, which make the system more efficient. The parameters obtained by fitting hyperbolae to hyperbolic signatures are very important features; they can be used to estimate the location and size of the related target objects and the average propagation velocity of the electromagnetic wave in the medium. The effectiveness of the proposed system is tested on both synthetic and real GPR data.

Ground-penetrating radar refraction imaging with stacked refraction convolution section method

We have evaluated a technique initially developed for the seismic refraction imaging, the stacked refraction convolution section (SRCS), which we have properly adapted to process ground-penetrating radar (GPR) refraction data. Through a mute operation, the subsurface refracting signals, recorded by the receiver from two reciprocal sources, are selected. Following that, a velocity analysis by means of the crosscorrelation of the refracted signals and the convolution of resulting traces is performed. The refraction image in intercept times is successively derived from three main steps, namely: (1) the convolution of the subsurface refracted signals, (2) the crosscorrelation of convolved trace with the reciprocal refracted signal, and (3) the stacking of crosscorrelated traces over all source couples. The technique is not only suitable for the processing of GPR data acquired with two or more reciprocal common source profiles but it is also convenient for its low acquisition cost in addition to the simplicity of software implementation and short processing times. We have evaluated the technique on a real GPR data set to characterize a near-surface morphostructure associated with a deep-seated gravitational slope deformation affecting Mt. Watles (Upper Venosta Valley, Italy). Results of the SRCS technique were validated against the direct trenching log data up to approximately 3 m in depth and complemented by the reflection processing outputs of common-source and common-offset data acquired along the line. The SRCS and common-midpoint processing provide the best reconstruction of the subsurface morphology of a shallow basement (approximately [Formula: see text] depth), characterized by a velocity range of [Formula: see text] and made of strongly to moderately weathered paragneiss. The full-wave modeling response of the reconstructed model demonstrates good agreement with the recorded signals.

An MM-Based Algorithm for -Regularized Least-Squares Estimation With an Application to Ground Penetrating Radar Image Reconstruction

An estimation method known as least absolute shrinkage and selection operator (LASSO) or ℓ1-regularized LS estimation has been found to perform well in a number of applications. In this paper, we use the majorize-minimize method to develop an algorithm for minimizing the LASSO objective function, which is the sum of a linear LS objective function plus an ℓ1 penalty term. The proposed algorithm, which we call the LASSO estimation via majorization-minimization (LMM) algorithm, is straightforward to implement, parallelizable, and guaranteed to produce LASSO objective function values that monotonically decrease. In addition, we formulate an extension of the LMM algorithm for reconstructing ground penetrating radar (GPR) images, that is much faster than the standard LMM algorithm and utilizes significantly less memory. Thus, the GPR specific LMM (GPR-LMM) algorithm is able to accommodate the big data associated with GPR imaging. We compare our proposed algorithms to the state-of-the-art ℓ1-regularized LS algorithms using a time and space complexity analysis. The GPR-LMM greatly outperforms the competing algorithms in terms of the performance metrics we considered. In addition, the reconstruction results of the standard LMM and GPR-LMM algorithms are evaluated using both simulated and real GPR data.

Attenuation analysis of real GPR wavelets: The equivalent amplitude spectrum (EAS)

Absorption of a Ground Penetrating Radar (GPR) pulse is a frequency dependent attenuation mechanism which causes a spectral shift on the dominant frequency of GPR data. Both energy variation of GPR amplitude spectrum and spectral shift were used for the estimation of Quality Factor (Q*) and subsequently the characterization of the subsurface material properties. The variation of the amplitude spectrum energy has been studied by Spectral Ratio (SR) method and the frequency shift by the estimation of the Frequency Centroid Shift (FCS) or the Frequency Peak Shift (FPS) methods. The FPS method is more automatic, less robust. This work aims to increase the robustness of the FPS method by fitting a part of the amplitude spectrum of GPR data with Ricker, Gaussian, Sigmoid–Gaussian or Ricker–Gaussian functions. These functions fit different parts of the spectrum of a GPR reference wavelet and the Equivalent Amplitude Spectrum (EAS) is selected, reproducing Q* values used in forward Q* modeling analysis. Then, only the peak frequencies and the time differences between the reference wavelet and the subsequent reflected wavelets are used to estimate Q*. As long as the EAS is estimated, it is used for Q* evaluation in all the GPR section, under the assumption that the selected reference wavelet is representative. De-phasing and constant phase shift, for obtaining symmetrical wavelets, proved useful in the sufficiency of the horizons picking. Synthetic, experimental and real GPR data were examined in order to demonstrate the effectiveness of the proposed methodology.

Automated reflection picking and polarity assessment through attribute analysis: Theory and application to synthetic and real ground-penetrating radar data

We have developed a procedure to automatically detect, pick, and characterize reflections in either seismic or ground-penetrating radar (GPR) data sets. Accurate picking results are essential in many research and application fields and are mandatory to recover the underground stratigraphy and reflectivity and, therefore, to create a more constrained subsurface model. We use the cosine of the instantaneous phase to track events with lateral phase continuity and record them in terms of time-space positions, amplitude, and polarity. Moreover, we use the cosine phase to reconstruct the shape of the reflected wavelets by averaging it along the picked horizons, to recover the initial phase of each reflection and, therefore, their polarities. By comparing the polarities of the transmitted and reflected wavelets, we can recover the impedance contrasts in the subsurface and evaluate the properties of the materials. We applied the picking and polarity assessment algorithms to a synthetic data set. We inverted the picked amplitudes and traveltimes to test the accuracy of the procedure by comparing the calculated stratigraphy and velocity distribution with the initial model. The inverted results were consistent with the original data, with most discordances in the stratigraphy within a few tens of centimeters, using a wavelet with a central frequency equal to 300 MHz. Local larger discrepancies were probably caused by interferences altering the amplitudes and resulting in the overestimation of the impedance contrasts. We also applied automatic picking and phase assessment procedures to real glaciological and archaeological GPR surveys, in which the influence of noise and diffractions was critically evaluated by comparing the picked results obtained from raw and processed data.

GPR Signal Denoising and Target Extraction With the CEEMD Method

In this letter, we apply a time and frequency analysis method based on the complete ensemble empirical mode decomposition (CEEMD) method in ground-penetrating radar (GPR) signal processing. It decomposes the GPR signal into a sum of oscillatory components, with guaranteed positive and smoothly varying instantaneous frequencies. The key idea of this method relies on averaging the modes obtained by empirical mode decomposition (EMD) applied to several realizations of Gaussian white noise added to the original signal. It can solve the mode-mixing problem in the EMD method and improve the resolution of ensemble EMD (EEMD) when the signal has a low signal-to-noise ratio. First, we analyze the difference between the basic theory of EMD, EEMD, and CEEMD. Then, we compare the time and frequency analysis with Hilbert–Huang transform to test the results of different methods. The synthetic and real GPR data demonstrate that CEEMD promises higher spectral–spatial resolution than the other two EMD methods in GPR signal denoising and target extraction. Its decomposition is complete, with a numerically negligible error.

Time‐varying band‐pass filtering GPR data by self‐inverse filtering

ABSTRACTEven though ground penetrating radar data signal processing has already been studied by many researchers, more research is needed and expected from automatic ground penetrating radar data analysis. An automatic band‐pass filtering procedure can lead to sufficient real‐time data interpretation as signal buried in noise could be amplified. Ground penetrating radar traces are highly non‐stationary, requiring time‐varying processing techniques. An algorithm, based on self‐inverse filtering, which is a special case of inverse filtering, was implemented. It is a ground penetrating radar trace filtering approach and is implemented by applying inverse filtering in each time sample in the time‐frequency domain. Applied on a synthetic trace, this algorithm performed better than a stationary band‐pass filter and empirical mode decomposition family methods, whereas its application on real ground penetrating radar data from two different sites enhanced reflections buried in noise without the need to test different high‐frequency band stops and with minimum distortion of the signal and the initial temporal resolution of the data.

Development of an Airborne Ground Penetrating Radar System: Antenna Design, Laboratory Experiment, and Numerical Simulation

A stepped-frequency continuous wave (SFCW) airborne ground penetrating radar (GPR) system was built using a vector network analyzer (VNA), an optical-to-electric ( <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">${\mmb O}\!/{\mmb E}$</tex> </formula> ) converter, an electric-to-optical ( <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">${\mmb E}/{\mmb O}$</tex></formula> ) converter, and the resistively loaded planer dipole antennas. The resistively loaded antennas are based on the Altshuler discrete resistive loading approach. The distance from the loading point to the feeding point and the resistance has been optimized; in addition, a balun at the feeding point was introduced to convert the unbalanced signal from the coax to balanced signal. Test results show that the manufactured antennas have a wide bandwidth frequency range from 55 to 275 MHz. A laboratory experiment using the developed system was conducted in a building with high ceilings. The results registered all basic features including the ground, sub-ground surface, and the lamp hanging from the ceiling. Numerical modeling by finite-difference time-domain (FDTD) method using two-dimensional (2-D) transverse electric (TE) verifies the presence of these registered features in the measured airborne data. A bedrock outcrop in ice sheet model is reconstructed based on a piece of real airborne GPR data, the modeling profile results show excellent consistent with the observed data. The capability of airborne GPR detection of subsurface features is clearly validated.

Characterization of collapsed paleocave systems using GPR attributes

Abstract Although the origin and evolution of karst collapsed paleocave systems are active topics of research, the characterization of the 3D internal geometry of these systems still requires further investigation. Ground penetrating radar (GPR) survey was performed to mapping subsurface collapsed paleocaves in carbonate sequence in the Cretaceous Potiguar Basin (northeastern Brazil), which can be considered analogous to hydrocarbon reservoir. To better identify and delineate the subsurface collapsed paleocaves several attributes, i.e. mathematical operators, such as Instantaneous Frequency, Amplitude, and attributes related to geometry of the reflectors, were applied to synthetic and real GPR data. The resulting radargrams helped the characterization of the electromagnetic response pattern of the collapsed paleocaves with respect to the host limestone and allowed an improved detection of the numerous fractures and clast contacts caused by the collapse of the ceiling and walls of the paleocaves. The GPR attributes also provide an enhanced three-dimensional imaging of the paleocave system internal geometry. Furthermore, changes in electrical impedance could be correlated with variations in porosity in the collapsed paleocaves and their bedrock. These results are consistent with petrological analysis of rock samples collected in the study area.

GPR identification of voids inside concrete based on the support vector machine algorithm

Voids inside reinforced concrete, which affect structural safety, are identified from ground penetrating radar (GPR) images using a completely automatic method based on the support vector machine (SVM) algorithm. The entire process can be characterized into four steps: (1) the original SVM model is built by training synthetic GPR data generated by finite difference time domain simulation and after data preprocessing, segmentation and feature extraction. (2) The classification accuracy of different kernel functions is compared with the cross-validation method and the penalty factor (c) of the SVM and the coefficient (σ2) of kernel functions are optimized by using the grid algorithm and the genetic algorithm. (3) To test the success of classification, this model is then verified and validated by applying it to another set of synthetic GPR data. The result shows a high success rate for classification. (4) This original classifier model is finally applied to a set of real GPR data to identify and classify voids. The result is less than ideal when compared with its application to synthetic data before the original model is improved. In general, this study shows that the SVM exhibits promising performance in the GPR identification of voids inside reinforced concrete. Nevertheless, the recognition of shape and distribution of voids may need further improvement.

지면 투과 레이더(GPR) 기반의 지뢰 탐지 시스템을 위한 표적 후보 검출 기법

최근 금속을 포함하여 다양한 재질의 지뢰를 탐지할 수 있는 지면 투과 레이더 (Ground Penetrating Radar, GPR)가 지뢰 탐지 분야에서 주목 받고 있으며 지면 투과 레이더를 이용한 지뢰 탐지 기술이 활발하게 연구되고 있다. 본 논문에서는 1차원 지면 투과 레이더 배열을 이용하여 지뢰가 존재할 가능성이 높은 후보 지점을 검출하는 방법을 제안하였다. 제안 기법은 먼저 지면 투과 레이더에 수신되는 신호 세기의 비선형적 감쇠 특성을 반영하여, 지연 시간에 따른 지면 투과 레이더 응답 신호의 세기를 효과적으로 나타낼 수 있는 회귀 모델을 추정하였다. 이를 토대로 지면 투과 레이더 응답 신호를 분석해보면, 토양 신호들은 주로 회귀 모델의 근처에 밀집되어 있고, 지뢰일 가능성이 있는 후보 신호들은 회귀 모델로부터 멀리 떨어져서 분포하는 특성을 보인다. 이러한 사실을 기반으로, 지면 투과 레이더의 측정값과 회귀 모델의 예측 값에 대한 예측 오차를 계산하여 토양 신호로부터 지뢰일 가능성이 있는 후보 신호들을 분리하였다. 후보 신호에는 잡음과 표적 신호가 섞여 있으므로 표적 신호가 갖는 고유한 기하학적 특성을 이용하여 최종적으로 표적 신호를 검출하였다. 성능 검증을 위해 실내 환경에서 획득한 지면 투과 레이더 데이터를 이용하여 실험을 수행한 결과, 제안하는 전처리 기법의 지뢰 탐지율 대비 오경보율이 낮은 것을 확인할 수 있었다.

Topographic migration of GPR data: Examples from Chad and Mongolia

Most ground-penetrating radar (GPR) measurements are performed on nearly flat areas. If strongly dipping reflections and/or diffractions are present in the GPR data, a classical migration-processing step is needed in order to determine the geometries of shallow structures. Nevertheless, a standard migration routine is not suitable for GPR data collected on areas showing a variable and large topographic relief. To take into account topographic variations, the GPR data are, in general, corrected by applying static shifts instead of using an appropriate topographic migration that would place the reflectors at their correct locations with the right dip angle. In this article, we present an overview of Kirchhoff's migration and show the importance of topographic migration in the case where the depth of the target structures is of the same order as the relief variations. Examples of synthetic and real GPR data are shown to illustrate the efficiency of the topographic migration.

An improved back projection imaging algorithm for subsurface target detection

An improved back projection imaging algorithm for subsurface target detection is presented in this paper. Firstly, the characteristic of the scattering data at each time-delay curve in the traditional back projection imaging procedure is analyzed. Secondly, a weight factor is designed for each focal point and an improved back projection imaging algorithm is presented. Thirdly, the simulation of the improved imaging algorithm is processed. The imaging results of both the simulation data and the real ground penetrating radar data show the effectiveness of this imaging algorithm.

Maxwell curl equation datuming for GPR based on the Kirchhoff integral solution and application in a tunnel grouting test

ABSTRACTIn two‐dimensional (2D) ground‐penetrating radar (GPR) data, the reflections from detection targets in depth can be severely obscured by strong scattering generated from near‐surface non‐target structures. By using GPR as a geotechnical non‐destructive testing device, it is still a challenge to eliminate the strong scattering caused by near‐surface rebars in a tunnel lining and to image and asses the grouting condition behind the tunnel lining. This study proposes a method for the reconstruction of GPR images, termed Maxwell curl equation datuming. To eliminate the deleterious effect caused by near‐surface diffractive scattering, we have redefined the reference surface to an actual geologic/engineered interface by using Maxwell curl equation datuming methodology based on the Kirchhoff integral solution. The datuming procedure can redefine the reference surface to a deeper horizon on which GPR transmitters and receivers appear to be located. We conducted a comparison between the datuming and migration to synthetic examples and case studies are presented for a physical model and real GPR data for assessments of shield tunnel grouting in the Shanghai Metro line No.9. The results show that the datuming technique is able to eliminate the strong scattering related to near‐surface rebars in a tunnel lining and improves the quality of deeper images beneath the tunnel lining. The datuming images make it easy to identify the distribution of tunnel grouting.

A Least Squares Approach to Buried Object Detection Using Ground Penetrating Radar

In this study, we utilize the least squares formulation to solve the real-time buried object detection problem. Least squares estimation is used to estimate the next ground penetrating radar (GPR) signal from previous samples, where there is no underground object. If the measured GPR signal is considerably different than the estimated signal, presence of an underground object is concluded. In order to attain real-time performance, Cholesky factorization is used when solving the linear systems. The proposed approach is tested on an extensive data set of different surrogate mines and other objects that are commonly encountered under the ground. The data are collected from three different terrains with different soil types to reveal the true performance of the method. It is demonstrated that our approach achieves almost 100% performance with a false alarm rate of approximately 10% on real GPR data collected with a handheld GPR system.

Visualizing Unsaturated Flow Phenomena Using High‐Resolution Reflection Ground Penetrating Radar

Unsaturated flow phenomena, such as unstable wetting fronts and preferential flow, cannot be investigated using small‐scale sampling. Dye tracer experiments can help visualize the dynamics of water flow but are destructive and therefore irreproducible. We investigated the applicability of high‐resolution ground penetrating radar (GPR) for nondestructive visualization of unsaturated flow patterns arising from a forced infiltration experiment. Synthetic studies using a reflection GPR two‐dimensional finite difference time domain modeling code indicated that differences in water content caused by preferential flow and fingering could be resolved. Moisture content contrasts down to approximately 2.5% within the top 2 m were detectable, but with increasing degrees of heterogeneity in the subsurface it becomes difficult to distinguish these moisture content changes. We conducted a field experiment in which 100 mm (900 L) of Brilliant Blue dyed water was infiltrated across a 3‐ by 3‐m area in relatively homogenous and undisturbed sandy alluvial sediments. Reflection GPR data were collected before and after infiltration. Dye‐staining patterns, revealed by excavating a 2‐m‐deep trench through the infiltration area, were compared with changes in the GPR data. Reflection amplitude changes as well as reflection delay revealed significant differences within the dye‐stained area. The GPR data provided information about the unsaturated flow below the extent of the dye staining, and the results of the synthetic GPR modeling, as well as the observed changes in the real GPR data set, underline the potential of reflection GPR as a nondestructive method to map unsaturated flow phenomena.

FOCUSING GROUND PENETRATING RADAR IMAGES WITH VERTICAL OFFSET FILTERING

Existing focusing techniques for Ground Penetrating Radar (GPR) rely on migration of 2D or 3D images to remove clutter originating from objects laterally ofiset from the antenna. In applications requiring real-time focusing, a method operating on 1D trace data is required. This paper presents a new algorithm for focusing GPR images, the Vertical Ofiset Filter (VOF), using simulated and real GPR data.