GPR - Science topic

Thank you all for your very useful answers. Video on the Jordanian shore of the Dead Sea . Najib Abou Karaki

I’m civil engineer in Japan. I have experienced maintenance of railroad structure for 40years.

We have been used GPR Inspection system on railroad maintenance for 30years.

It works underground 2 or 3 meters limit depth.

It depends on the geologic characteristics. It interfered performance of GPR by alluvium,

Especially, clay interfere communication of electric wave. Moreover, it deflect electric wave by rail and concrete pillow.

If road bet of soil condition is alluvium, performance of GPR will be same.

I think.

You may use commercial software such as Reflexw to output each channel of data from 3D GPR. Then a mesh code can help you generate the C-Scan, D-Scan or any data slice from the 3D GPR data.

Hope this helps! Feel free to exchange more technical details via siqi2wang@polyu.edu.hk.

Agreeding with Doug, use seismic methods to evaluate soil stiffness or strength. MASW, or crosshole/downhole tests (if boreholes available).

Prof. Ibrar Iqbal

This paper can give some insight or probably more exact references, I only noticed that they used a refraction technique and not a reflection as is well known in seismics.

I did not know before about that interesting seismoelectric effect you mention in your thread.

Best Regards.

You can find your answers in my recently published article. https://www.researchgate.net/publication/354354664_Improving_the_Resolution_of_GRACE_Data_for_Spatio-Temporal_Groundwater_Storage_Assessment

This is just history, and how things are measured. It is also not true that microwaves are only referred to by frequency. They are called centimetre waves, and 100 GHz is referred to as mm wave. Even at low frequency, am radio at 200 kHz was known as long wave, and we looked for it at 1500 metres. Above that was short-wave! At low frequencies it is easier to measure the time of a period, or how many periods there are in a second. With light it is easier to measure the wavelength using diffraction by a grating (which has its line spacing measured in microns, or lines per cm), with a formula that relates angle to wavelength and grating period.

I am learning to think in both, because I am a microwave man, but am now working in near infra red, so am just learning what that is in frequency, but as I type this I find I still have to work it out and don't know wavelength and frequency instinctively yet, as I do from 1 GHz to 300 GHz.

Please keep in mind that you will not be able to derive seismic velocities from GPR measurements. If you are however interested in imaging the structure of the soil and topmost layers, then GPR will offer you highest imaging resolution.

Each of these software has advantages and disadvantages, but in general, they are not significantly different. I personally prefer the MATGPR package because it is free and can be easily used.

i am using regression app in matlab.

You can try gprMax (www.gprMax .com) which is a Free Open Source FDTD solver used to model GPR by many researchers for many years. There is no GUI to build models though if this is what you need.

Hi Filippo, I am not aware of any GPR pulse speed slower than 33 cm/ns. The table that you show does not list a speed of "0.030" m/ns, but "0.03" m/ns. So, I think we are dealing with the effect of rounding inaccuracy rather than an actually observed velocity of 30 cm/ns. I would like to recommend Jürg Leckbusch's site on GPR antenna and material parameters: http://gpr-parameters.ch/material_properties.php

Thank you very much for this very useful gloassary! Why are the terms "section" and "profile" missing? You have the term "cross section" . I am missing as well the terms "cross-line" (you have "X line") and "in-line" to describe different samle spacings. I have the feeling that the terms "A-scan", "B-scan" and "C-scan" (for trace, section/profile, slice) are somewhat outdated. What do you think?

I believe that the only ways to improve the penetration of EM waves through high-saline soils are:

(i) use lower frequencies

(ii) increase the transmitted power

(iii) both of the above.

Regards,

Ray

I think manual set up using TerraSirch mode is always better for the cavity search.

Samples can conventiently be held in a 2D array of samples within each FM sweep (rows) vs samples from successive sweeps (columns). You will first want to focus the array by making phase offsets on samples as a function of their range and location within the synthetic array. Then a 2D FFT process of data will yield the cross-range vs range map. The FFT of the slow-time samples from successive sweeps gives the Doppler shift of a point which is a function of its cross-range location. The FFT of the fast-time samples within any given sweep gives the beat frequency which is a function of its range. This will get you a basic image/map.

Dear Konstantinos, please, try this article by Kovin and Anderson, available at RG website.

Basically, the idea is use the same principles such as 3D seismic interpretation: it means: time-slices, horizon-slices, and attributes, e.g., coherency, curvature, spectral decomposition, to identified the fractures.

Please, let me know, I can send to you an example from seismic interpretation.

Mario

When you import the data into ReflexW (that is when you convert it to the internally used format), use a Geometry file (e.g. "Geometry.1") in which you have described the geometry of the survey (see page 211 in the ReflexW manual for the format of the geometry file: https://www.sandmeier-geo.de/Download/reflexw_manual_a4.pdf ). In the import menu select "geometry file" under "ConversionMode". Then process the 2D profiles. Once you are happy with a representative profile, save the applied processing sequence as batch job and apply it through "Sequence Processing" to all profiles. Then exit the 2D mode and enter the 3D module. Here you have to specify for the X- and Y-Rasterincrement the crossline and inline sample spacing, so 2.5m and whatever your inline trace spacing was. The X- and YInterpolation are the horizontal dimensions of the generated and interpolated 3D volume. When ReflexW complains about too many data values, then increase the X- and YInterpolation values. It is not meaningful to interpolate to 10cm cell size when your profile spacing was 250cm. Maybe try 50cm and 10cm. for crossline and inline spacing?

Hi,

As for methodological approach, I am dealing with the similar kind of research, so I would like to advise you using either 1) interrupted time series (ITS) analysis (described i.e. at “Experimental and quasi-experimental designs for generalized causal inference” by W. R. Shadish et al., 2002), or, if you prefer qualitative methods, 2) Longitudinal case selection strategy, which is kind of ITS analysis (from “Selecting Cases for Intensive Analysis: A Diversity of Goals and Methods by John Gerring and Lee Cojocaru, 2016).

I'm glad to attend, I will get in touch with you.

Dear Ms Marshal,

Less resistive clays rich horizon over more resistive sedimentary layer can limit the GPR investigation depth. Clayey layers are conductive and attenuate the radar signals. GPR is not appropriate in areas where the surface layer resistivity is less than 100 Ohm. m, that is the case of most clay rich soil horizons. Regards, Anicet

The answer depends on the depth of target. As you know very well depth penetration of GPR method is very limited. Besides, deriving the fractures within rocks based on GPS is very hard. Therefore, classification of rock type beneath a site have to be done using other geophysical methods.

There are lots of software out there as discussed, depending on what you want to do with your data and ultimately your budget and what system you are using to collect the data. Most software will take all formats but please be aware that the dual frequency antennas are not universally accepted in all software.

Radan 7 is good for 1d, 2d and 3d data as previously mentioned.

Gred HD from IDS is good but can be a little memory heavy on the pc.

GPRSoft is good for 2d and pro version is good with 3d data

ReflexW is OK too as you know.

GPRSlice has a comprehensive module system for all applications.

The list goes on, data interpretation is something else, there are many sites on the internet and all give good information so a good google search and afternoon of research can narrow things down a lot for specifics. Start with equipment manufacturers sites and get the basics from them then see where you go from there.

Have fun.

From the SignalPrediction server http://www.cbs.dtu.dk/services/SignalP/ :

"NOTE (added May 2018): Remember, the presence or absence of a signal peptide is not the whole story about the localization of a protein! If you want to find out more about the sorting of your eukaryotic proteins, try the protein subcellular localization predictor DeepLoc."

and from the Wikipedia article on signal peptide https://en.wikipedia.org/wiki/Signal_peptide:

"Although most type I membrane-bound proteins have signal peptides, the majority of type II and multi-spanning membrane-bound proteins are targeted to the secretory pathway by their first transmembrane domain, which biochemically resembles a signal sequence except that it is not cleaved."

Do You have the 3D data interpretation module for Reflexw? This is normally supplied as standard.

Thank you for bringing this up! Regarding the statement "Users must be of a class of professionals as designated on the labelling on the GPR instrument", which professionals qualify as GPR users and how is "professionals" defined in this context (someone who earns a living by using GPR, someone with a certain degree, education, training)?

You need to check the amplitude of different frequency in processing software and based on the amplitude of frequency you can select any frequency is suitable for your data.

Peter

The general issue that you pose for the GPR technology is generalizable for other ones in geophysics. (see for example the use and applications of seismic tomography)

It Is the general issue of "the mature technologies" and we should to satisfied with the fact that GPR technology became a mature technology.

A mature technology is characterized by:

-consolidated basic theoretical standards

-relatively low instrumental costs

-relatively low logistic/operational costs

-relatively low software costs

-methodological simplicity (low processing time with not high analyst level request)

-high demand for this technology in different application fields

As a consequence, a mature technology has a high diffusion inside applied science operators and professionals and it opens opportunities for different stakeholders (mainly economic-professional) as you claimed in the blog. It is good.

I agree with you that we must spend more effort to diffuse and improve in all forums (educational, professional and societal-economic) the best standards and the best practices of the technology. But Its not matter to defend a mature technology in regards the "absolute quality" because, in this case, it is regulated only by the "natural selection" mainly operated by "the final users of results" (in scientific and professional frameworks). It is so, even though this could appear "raw". Differently should be a matter to discuss the technological-scientific future of a mature technology, in Research&Development, (as it is GPR), because in this case, the effort to be made is far greater than that of an unripe technology. In my opinion, when a technology goes forward, in a sustainable way, it also improves the feedback on its applications.

I recommend the classic matGPR R3

Hi Mimi. It would be better to share what you have tried. Then it would be easy to solve your problem.

However, over the top of my head, I think you need to execute the code by using the Tx-rx steps and repeat it for N times to get a B or C-scan

Thanks very much Dr. Omid Ahmadi for sharing important informations

Regards!

You don't need any paper refernce to answer these questions. Actually these are the most basic knowledge on GPR that one may want to have. These information can be found in all basic books, such as Daniels. I'd also better say that you can find all these in any signal related book, website. Simply google, for example: "Band pass filter theory"

Hi Roger,

Recently geoelectric tomography and ground penetrating radar is widely used in the industry to detect underground voids. Both techniques are based on physical contrast between void and surrounding soil. As your aim is to scan for 15-40m depth, the results will have significant noise, as changes in the ground material at layer boundaries cause transmitted signal to be reflected in a different manner and the energy of transmitted wave is absorbed by ground material. This absorption is smaller for material such as granite, dry sand, snow and water. Therefore, GPR will penetrate much deeper than materials like clay which absorbs more electromagnetic energy. However, most of the the time your interested zone is well below the ground water table. Therefore, the disturbance is higher.

The best thing is talk to some industrial experts. They may have some modern techniques. I have attached few pages from the following reference for more details about the general background. Hope it might help you. Good luck

Take a look at this FDTD course that includes videos, lecture notes, coding sessions, and other resources to help you develop your own codes.

I agree with the previous comments that it would be most appropriate to perform multi-offset GPR measurements at the site of interest in order to determine the radar wave velocity. This ensure that the same volume of subsurface is being investigated.

If you find reference in situ permittivity measurements essential and TDR is not an option, a relatively cheap way may be to use capacitance probes or water content sensors based on a ring oscillator. Some of the more advanced dielectric soil moisture sensors are factory-calibrated and are able to provide relatively accurate permittivity estimates. More information can be obtained here:

Qu, W., H.R. Bogena, J.A. Huisman and H. Vereecken. 2013. Calibration of a novel low-cost time domain transmission soil water content sensor. Vadose Zone Journal, 12(1), doi:10.2136/vzj2012.0139.

Bogena, H., J.A. Huisman, C. Oberdoerster and H. Vereecken. 2007. Evaluation of a low-cost soil water content sensor for wireless network applications. Journal of Hydrology, 344, 32-42.

I am sure that there are more sensors that can do this. Please do realize that TDR will likely provide more accurate permittivity estimates. In addition, the sampling volume of these in-situ sensors is much smaller, which may lead to interpretation problems.

Hope this helps!

Best,

Sander

To display 3D GPR diagrams you can

1) import your data into OpendTect and display them with OpendTect. "OpendTect is a free, open source seismic interpretation system and software development platform". See https://www.dgbes.com

2) import your data into R (programming environment similar to Matlab) and display them in 3D using the RGL package. To do that, you can use the R-package RGPR (https://github.com/emanuelhuber/RGPR) that I developed. RGPR supports the following binary formats: Sensors & Software file format (.DT1, .HD), MALA file format (.rd3, .rad), SEG-Y file format (.sgy) from RadSys Zond GPR device. See this tutorial to learn how to read and plot 3D GPR data: http://emanuelhuber.github.io/RGPR/RGPR_tutorial_RGPR-survey.html

The R language (https://cran.r-project.org/) is an alternative to MATLAB or python.

See for example the R-package RGPR that I developed:

RGPR is R package to visualize, analyze, process and interpret ground-penetrating radar (GPR) data. It can read .dt1, .rd3 and .segy GPR data file formats, plot GPR data in 1D, 2D, and 3D. There are not that many processing functions but because this is an on going project, more functions will be added in a near future.

Dear Saeid,

you are right, so i want the standard values of the GPR antenna parameters to take it into my considerations.

Unqualified comment: What is always ( within existing scientific paradigms )  in any kind of investigation in fact or essentially (at times unknowingly)  desired or  intended  is quantification (an arbitrary ) of gravities and/or simple or uncomplicated manifestations thereof.  Gravity variation introduces or manifests as variation in almost every thing. Stress variation is one uncomplicated (setting aside thermodynamics) manifestation of gravity variation. 

GPR is very mature method of locating cables, pipes and other buried features. It does have its limitations and is not always 100% effective. For example, it does not work well in moist heavy clay soils. Works best in dry sandy soil. It is difficult to scan below reinforced concrete or in stabilized material. It has limited depth of penetration and resolution. Lower frequency antennas have greater penetration but lower resolution. Therefore, there is always a trade off between depth and resolution. Smaller services are difficult to detect. Plastic services are more difficult to detect. Sometimes disturbed ground and trench boundaries can be detected even though the service is not detected.

There is a wide variety of GPR antennas and equipment. GPR equipment is not cheap  and can cost anywhere from USD15k to 50k. Some common brands include Easy Radar, GSSI, IDS, Mala, Radar Team, Sensors & Software. For basic training you could get going with just one day but more would be required to be proficient.

Every GPR system contains one or multiple antennas. These antennas transmit and receive radio frequency waves. 

GPR has the same basic principles as a metal detector. A metal detector sends energy into the earth in up to 17 frequencies. When that energy meets a metallic object, it is translated into a recognizable tone. The GPR sends out millions of frequencies that return to the antenna and translate material composition definition in the subsurface. Radar is sensitive to changes in material composition. 

Radar GPR Systems are designed to display differences in material composition. They can be used to locate any object that has a different composition than it's surrounding materials. For example, water, a PVC pipe will have a different composition than the surrounding soil. Voids and excavations that have been filled in will also have different compositions than the surrounding soil.

The depth of your findings will be determined by three factors:

Soil type

Antenna frequency

Size of Target

you my take alook on the attached article. I hope it will help

Dear Dr. Igo,

Yes, GPR can. There are many publications regarding detection of subsurface discontinuities and stratification using GPR, however, as answered by the colleagues here, it is a matter of using the most suitable antenna that satisfies your target (resolution and penetration depth). You may have a look to this paper, for example:

Dear Gamil and dear all, please note that there more to GPR than the monostatic configuration you all appear to have in mind. So, to begin with the original question is improperly posed and could be recast as follows:

1) "Can we distinguish in the geometry of the target buried when scanning by MONOSTATIC GPR?". The answer is that any type of horizontal or subhorizontal interface can be resolved to the detail affordable by combination of host material, depth and size of the target on one hand, and the antenna frequency on the other. Right combinations allow for arbitrarilly good resolution. Conversely, vertical and subvertical interfaces cannot be resolved because they do not reflect the signal.

2) "Can we distinguish in the geometry of the target buried when scanning by BISTATIC GPR, say one used a densely spaced Common MidPoint mode?" If the combination of material, depth, size and antenna allows, the answer is generally yes we can, because in this case we receive returns from vertical and subvertical interfaces as well. 

3) "Can we distinguish in the geometry of the target buried when scanning in tomographic mode using antenna arrays?". If the target is smaller than the array and the combination of material, depth and antennae allows, the answer is a definite yes.  

A few forward modelling and migration exercises, preferably with FDTD and 2-D migration techniques can easily demonstrate these points.

Wish you all the best

Hi Shreedhar,

an overview and helpful references on feature selection is given in the article by Liu and Yu: "Toward Integrating Feature Selection Algorithms for Classification and Clustering".

If you want optimal features for your SVM you could use a wrapper method (the term is used in the article from before).  However, when using wrappers overfitting again is a serious issue (see Reunanens work entitled "Overfitting in feature selection: Pitfalls and solutions.")

Greetings, Sebastian

We have observed some areas having perfect dry sand with really high velocities, 0.18 - 0.185 m/ns. 

If this is rubble, then the location and depth give you good information as to where and how deep to dig.  You have three suspected targets out of a traverse of 120 metres six metres deep.  You will have saved a heck of a lot of digging by knowing where to dig.

If this is rubble, then this is likely a rescue mission or a bomb disposal/retrieval mission. Why do you need the computer to determine the nature of the target?  Shouldn't determining the presence and localization of an anomaly be enough?

Dewow in Reflex acts on each trace in the profile in turn. For each point on each trace, it calculates the mean value of all the points either side of that point within the time range specified; and subtracts this mean value from the original point. It does this as a running average through all the points in the trace, moving the time window with the point, It does this on each trace independantly.

I don't use this function very often as there are more efficient methods. All depends what you are trying to achieve. In my experience the fewer processing steps you can get away with, the better. Each processing step adds its own artifacts to the mix and you need to be aware of this.  

Hi;

Below are some references I would like to share. Please check GPR radargrams included in the references, especially interpreted radargrams. I think they could give some ideas that will help in your work, Sir.

In addition to the basic data processing steps such as background removal, I would like to suggest you to apply de-convolution and migration to reduce your data. It seems to me both data processing steps would make your data more interpretable.

I hope the references given below could be helpful

Kind Regards

REFERENCES

 

Xu, X., Xia, T., Venkatachalam, A., 2012. Development of high-speed ultrawideband ground-penetrating radar for rebar detection. J. Eng. Mech. 139, 272–285. doi:http://dx.doi.org/10.1061/(ASCE)EM.1943-7889.0000458

Yehia, S., Abudayyeh, O., Abdel-Qader, I., Zalt, A., 2008. Ground-Penetrating Radar, Chain Drag, and Ground Truth: Correlation of Bridge Deck Assessment Data. Transp. Res. Rec. J. Transp. Res. Board 2044, 39–50. doi:10.3141/2044-05

URL 1 https://fhwaapps.fhwa.dot.gov/ndep/DisplayTechnology.aspx?tech_id=24

URL 2 http://www.ijser.org/paper/Thickness-evaluation-of-asphalt-and-base-layers-of-some-major-and-minor-arterial-roads-in-Kumasi-using-ground-penetrating-radar.html

URL 3 http://foundationperformance.org/pastpresentations/gehrig_paper_march2004.pdf

URL 4 http://ntl.bts.gov/lib/43000/43000/43073/134431_FR.pdf

URL 5 http://www.sciencedirect.com/science/article/pii/S0963869504001021

URL 6 http://www.mdpi.com/1424-8220/9/3/1454/htm

I want it @Farad Gholamian

Thank you.

Hello,

I'd say "no way" - giving you a link to a commercial vendor of antennas - just to show you what's currently sold (and considered to be applicable):

These guys have all types of antennas. You are free to surf a bit and find out what's offered for your frequency range.

A very short glimpse brought up magnetic loop antennas for the lowest frequencies...

And if it comes to efficiency, you may want to consider log-periodic antennas instead of the bowtie (or biconical) antenna.

Regards

Hi, I also use Reflex-W.  With this programme I would be able to help you.

You might look at our use of UAV-based thermal imaging described in the following article: 

I would add to Christoph's answers (which are great by the way):

- better automated target recognition software

Presumably they are mainly vertical, in which case you could try refraction seismic.  Use a drop weight or hammer and plate and move an array of receiver stations around. Compare cavity wall sound insulation.  I did something similar on a limestone quarry top that was cut by small faults; I can scan the resulting map for you if you wish.

GPR is a Microsoft Excel based algorithm for Global Pattern Recognition (GPR) for PCR/qRT-PCR based techniques. If your GPR fold change for a gene is 96.45, means that the said gene is expressed at 96 fold higher level (compared to control).

Basically you are measuring deferentially expressed genes from multiple replicates ( from at least two conditions/samples).

See some of the links below:

https://www.bhbio.com/BHB/(S(pbw4hguf2capymwtwa13fam5))/GUI/SP/Gpr.aspx?AspxAutoDetectCookieSupport=1

Hope this information is helpful.

I would be very surprise if there is a relationship between dielectric constant and pavement strength. Moisture and density and compositional changes in the unbound material have a larger affect on dielectric than any related strength variation in the material itself. Dielectric can be used to map variations in pavement materials and possibly by inference information regarding its condition. Particularly when considered in context with thickness variation and other signal anomalies.  Variations in dielectric can be determine using wave a ratio analysis, provided that sufficient reflection amplitude can be achieved from the material layers in question. Some relavent papers attached.

Ground Penetrating Radar Surveys to Characterize Pavement Layer Thickness Variations at GPS Sites  Kenneth R.Mase 1994

Road evaluation with ground penetrating radar

T Saarenketo, T Scullion - Journal of applied geophysics, 2000

Most GPR data formats simply store the binary floating point values one after the other, trace after trace. Hence if you know the number of samples per trace and desired trace number you can calculate the offset. Some formats also have a header that is easy to find when looking at the binary data. If it is a Mala dataset the header is in a separate file anyway. If there is a Segy output you'll find the documentation of the format in the internet.

Regards, Juerg Leckebusch

Bonjour,

les valeurs disponibles sont en effet généralement limitées à 1 GHz car sur les sondes co-axiales communément utilisées cela correspond généralement à la fréquence de coupure à partir de laquelle on n'a plus de mode TEM mais des modes complexes pour lesquels les modèles ne sont plus valables. Vous pourriez toutefois contacter Xavier Dérobert à l'IFSTTAR de Nantes car ils ont développé des sondes  qui permettent de travailler jusqu'à au moins 5 GHz de mémoire. Ensuite il faut quand même reconnaître qu'avec les antennes GSSI à 1.5 GHz dans l'air, dans le béton on est plus près du GHz. Donc ces résultats suffisent. Ensuite la dispersion électromagnétique si tant est qu'elle existe est faible à ces fréquences là.

Bonne journée et bien cordialement

this might help, im not sure

Thank you very mach for your intervention. I already had GPR data from an USA researcher .

If it is possible, I would like to know if you work on the GPR data. If so, I have a small question in the field of Clutter reduction from the Bscan data signals. This task is easy to do with simple method or not?.

Thank you again.