0% found this document useful (0 votes)
63 views8 pages

The Development of Ground Penetrating Radar (GPR) Data Processing

Uploaded by

Jayesh Sharma
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
63 views8 pages

The Development of Ground Penetrating Radar (GPR) Data Processing

Uploaded by

Jayesh Sharma
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 8

International Journal of Machine Learning and Computing, Vol. 9, No.

6, December 2019

The Development of Ground Penetrating Radar (GPR)


Data Processing
Baso Maruddani and Efri Sandi

interpretation of the GPR readings results [3]. Several


Abstract—Ground Penetrating Radar (GPR) is one of radar studies have been conducted in [4]-[6] to improve the
types that is often used to determine conditions inside or below accuracy of the GPR reading results system and the
some surface. GPR is also commonly used as a material development of the antenna model used. The antenna design
evaluation tool by its trait as a non-destructive testing (NDT).
One of the most important sections of GPR is signal processing
has also been compared comprehensively in [7]. Explained
system or GPR data processing that will filter all of the GPR in [7] that the antenna working frequency, antenna gain and
survey results. Reflection signals gained by the radar antenna antenna impedance were tested to determine the
are then filtered to discover any objects located below the compatibility with the GPR application that was
ground surface. The better the process in filtering data, the implemented.
more accurate the GPR interprets the gained signal. This study The research on GPR data processing has been conducted
aims to design a GPR data processing system that is sufficiently
be able to interpret the gained signal so that it can accurately in [2], [8]-[13]. In [2], software for GPR data processing
discover any objects located on the ground. This GPR data from R – language was created, called RGPR. RGPR is built
processing system is expected to work on a variety of based on two classes to filter and visualize GPR data by
frequencies and the application can be developed in various paying attention to step by step data filtering process. There
types. The aims of this study are designing and modifying the are many basic methods of GPR data processing which are
GPR data processing system. implemented in the RGPR. However, the RGPR is still
under development and improvement at this present. One of
Index Terms—GPR, processing, dewow, filtering, gain.
the methods that used in RGPR is as stated in [14].

I. INTRODUCTION
II. THE PRINCIPLES AND SIGNAL MEASUREMENT OF GPR
Applications for wireless communication systems vary
greatly. Radar is one of the applications where the Every material has electrical properties which those
transmitter and receiver are in the same place. Ground properties are important for knowing. Those properties are
Penetrating Radar (GPR) is a wireless communication permittivity (ε), conductivity (σ) and permeability (µ). In
system that is used to view any objects located under the GPR application, value of permittivity and conductivity is
ground surface or behind the walls. GPR can "view" through important while permeability is rare of concern. GPR is
the ground by sending electromagnetic signals to a certain most useful if used in low electrical loss materials. If
object and then gaining a reflection signal from the signal conductivity (σ) is zero, then GPR can be see under surface
sent earlier and reconstructing that reflected signal. deeper. In reality, low electrical loss materials are not
Reflected signals is reconstructed by processing the prevalent. Water under surface or moist material or clay rich
reflected signals to be an image. GPR design highly depends environment makes the capability of GPR to penetrate the
on the application used. Therefore, GPR to "view" the inner deeper surface is limited. Ground materials are composites
surface is basically different from GPR to observe the from many materials. Ice and water are examples that
shallow surface [1].1 represent a most case.
GPR has very wide applications, including for geology, In practice, the problem of low electricity losses is not
archeology, civil engineering, non-destructive testing (NDT) uncommon. An environment that rich in groundwater or
and in the field of defense and security. Several studies have saline can create conditions where GPR signal penetration is
been conducted on GPR. Both from the hardware side, very limited. Earth material is a combination of many other
namely the antenna, and from the software side [2]. On the ingredients or components. Water and ice represent several
hardware side, research that has been conducted was about cases where one main component exists. Simple beach sand
the selection of the GPR working frequency. The choice of is a mixture of soil grains, air, air, and dissolves in the air.
antenna working frequency used in GPR devices depends on Ground grains will usually be placed 60-80% of the
the application used, because each antenna has a radiation available volume. Understanding the physical nature of the
pattern and some characteristics will affect the final mixture is a key factor in the interpretation of the GPR
response. A mixture of materials that rarely shows
properties proportional to the volume fraction of its
Manuscript received July 17, 2019; revised October 15, 2019. This constituent components. In many cases, this can make
research is partially supported by Institute of Research and Community quantitative analysis impossible without additional
Service, Universitas Negeri Jakarta and DJA Institute Jakarta, Indonesia.
Baso Maruddani and Efri Sandi are with the Faculty of Engineering, information.
Univeritas Negeri Jakarta, DKI Jakarta, Indonesia. Baso Maruddani is also How GPR works is conceptually simple. Its aim is to
with DJA Institute, DKI Jakarta, Indonesia (e-mail: measure the amplitude of signal and time after excitation.
[email protected], [email protected]).

doi: 10.18178/ijmlc.2019.9.6.871 768


International Journal of Machine Learning and Computing, Vol. 9, No. 6, December 2019

The core of the GPR system is a time management unit transmitting. Transmitter blanking occurs when the direct-
which regulates the signal generating and detects signal signal from the transmitting antenna that propagates to the
echo as shown in Fig. 1. Most of GPR is works in time receiving antenna overlaps with reflected signals from
domain so that to synthesize time domain responses, objects below the surface
frequency domain is used. This is bandwidth and dynamic range problem. The
transmitter source usually emits a very large signal, and if
the receiver is near the cassing transmitter in GPR, the
receiver will see a very large direct signal transmitted. If this
signal is large enough, the electronic receiver will be excess
and will not receive the reflected signal. The duration of
signal propagation time changes inversely with bandwidth.
Resolution length ∆r related to bandwith is

III. THE CONCEPT OF GPR DATA PROCESSING


DEVELOPMENT
This study aims to develop a GPR data processing, which
is expected to be used freely. The development phase of
Fig. 1. Block diagram of GPR [3]. GPR data processing is briefly explained in the Fig. 4.

Characterization of radar systems is a complex assigment


because there are many problems that affect the operation
and use of the system. Electronic instrumentation factors
that govern the characterization of GPR are signal
generation, signal processsing, signal capture methods,
dynamic range, performance factors, center frequency and
bandwidth, support, and portability. The antenna converts
the electrical signal from and to the electromagnetic field.

Fig. 4. GPR data processing and analysis steps as explain in [3].


Fig. 2. Transmitter blanking that occurs because overlapping between direct
signal and reflected signal [3].
As explained in [3], the editing and rubber-banding
phases are the stages of preparing survey GPR data that will
be processed. Information about the frequency used, the
distance between lines, the number of lines, the number of
traces, the type of file used and the others are things that
must be known from the beginning so that the data
processing in the next step is not wrong. The dewow phase
is the stage that aims to eliminate the low frequency and DC
bias in the data. Wow is a noise that has a very low
frequency value, this occurs due to an electronic instrument
Fig. 3. Transmitter blanking that occurs because overlapping between two saturated by the value of large amplitude of direct waves and
reflected signals that have a similar path length [3]. air waves. Dewowing is a vital step as it reduces the data to
a mean zero level and, therefore, allows positive–negative
The GPR system has to record the data with an accuration colour filling to be used in the recorded traces [3]. If done
of time which less than 10 ps for 10,000 ns duration [3]. with incorrectly way, the data will be decayed and the low
Bandwidth measurement depends on the application and is frequency component will distort the spectrum from trace
directly related to resolution. The resolution of two topics data. This incorrectly processed data will affect to the next
related to "transmitter blanking" and "target separation" is data processing stage. If this process is done manually step
illustrated in Fig. 2 and Fig. 3. by step, it is better to eliminate the DC component first and
The transmitter blanking is caused by the inability of the then implement the filter. The dewow concept can be seen in
receiver to detect the signal until the transmitter has finished Fig. 5.

769
International Journal of Machine Learning and Computing, Vol. 9, No. 6, December 2019

TABLE I: APPROXIMATE DIELECTRIC PERMITTIVITIES, ELECTRICAL


CONDUCTIVITIES AND RADIOWAVE VELOCITIES FOR VARIOUS MATERIALS
Material Relative Conductivity Average
Permittivity (mS/m) Velocity (m/ns)
Air 1 0 0.3
Fresh Water 80 0.5 0.033
Sea Water 80 3000 0.01
Ice 3-4 0.01 0.16
Dry Sand 3-5 0.01 0.15
Saturated Sand 20-30 0.1-1 0.06
Limestone 4-8 0.5-2 0.12
Shales 5-15 1-100 0.09
Silts 5-30 1-100 0.07
Clays 5-40 2-1000 0.06
Granite 4-6 0.01-1 0.13
Anhydrites 3-4 0.01-1 0.13

Line 0 4
x 10
0
Fig. 5. Dewow concept stage for processing GPR data [3]. 3

50
2

The next phase is time zero correction. This aimed to 100

two way travel time (ns)


correct the initial time to be exactly the same with the 1
150
surface of the ground. Thermal drift, electronic instability,
200
cable length differences and variations in antenna airgap can 0

cause ‘jumps’ in the air/ground wavelet first arrival time [3]. 250 -1

The next phase is filtering, which is done by implementing 300


1D and 2D filtering to improve signal to noise ratio and -2

350
correcting image quality. There are many types of filters that -3

400
can be applied at this phase, from simple band-pass filters to 0 2 4 6 8 10
transform filters. Simple filters sometimes effective to position (meter)
Fig. 6. Image profile of GPR data before being processed for the 0th row
eliminate the high frequency noise and low frequencies. And data.
sometimes some complex filters are not needed. In general,
1D Plot Line 0 Trace 1
filters can be divided into two types, which are temporal 4
x 10
4

(one by one filtering the trace data) and spatial (filter for
3

some trace data). Examples of temporal filters are simple


mean, simple median, low pass, high pass and band pass. 2
signal amplitude (mV)

Examples of spatial filters are simple running average, 1

average subtraction, both background removal and spatial


low, high and bandpass filter. 0

The next stage is deconvolution. Deconvolution aims to -1

contrast signals for increasing the resolution by compressing


-2
GPR wavelet data into a narrow band. The purpose is to
remove the effect of the source wavelet from the GPR data -3

and leave the impulse response to subsurface layers. This -4


0 100 200 300 400 500 600 700 800 900 1000
deconvolution process assumes that the subsurface is two way travetime (ns)
horizontally layered and has a uniform intra-layer velocity Fig. 7. 1D plot signal trace.
and the signal reflection appears an interface coherently.
The next step is velocity analysis. The purpose of this
velocity analysis is to analyze the speed of the GPR signal. IV. DESIGN AND DISCUSSION OF GPR DATA PROCESSING
Electromagnetic signals are propagating in a medium which PROGRAM
depend on the characteristics of the media itself. If in the air,
the velocity of electromagnetic waves is 3.108 m/s. If other A. Data Preparation Process
than air, the velocity of the electromagnetic wave is smaller At the first time, the data processing program must be
than that value. Table I summarizes the electromagnetic initiated first. GPR data must be stored in a specified folder
wave velocity on certain media. so that the program created can directly accesses to the data
The next step is elevation correction and migration. The in the previously defined folder. The data used in this study
purpose of this step is to correct the effects of topography are data in [2]. The format of existing data that will be
and the effects of geometry survey also the effects of the processed is a data in matrix format which the data
spatial distribution energy. The next step is depth conversion. characteristics defined in the program are:
This step synchronizes between two-way travel time and • frequency
depth. As explained in the Table I, the depth of a surface can • the distance between the transmitter and the receiver
be converted from electromagnetic wave velocity and two- antenna
way travel time. • number of rows and number of trace

770
International Journal of Machine Learning and Computing, Vol. 9, No. 6, December 2019

• distance between lines (in meters) process is dewowing which aims to eliminate noise and DC
• GPR data storage folder bias components. Fig. 9 presents the GPR data profile after
• GPR post processing data storage folder dewowing process. Fig. 10 and Fig. 11 successively presents
There is a survey parameter which plays an important role a 1D plot for the signal trace after dewowing process and the
in this program. For each data that will be processed or signal amplitude ratio before and after dewowing process.
filtered, this survey parameter must be arranged as Line 0 4
x 10

organized as possible to be able to process the GPR data. On 0


4

the GPR data processing program that has been created 50 3


before, the bottom line is defined at the beginning of the
survey parameters on the bottom row. The survey parameter

two way travel time (ns)


100 2

is set to call GPR data in Line00.DT1 format. And then, the 1

150
program that has been created will call the other GPR data
0

files. The GPR data which tested in this program is a data 200
with four lines of GPR data. Then from the tested GPR data -1

profile, the distance for each rows is 0.2 meters. The 250 -2

distance between the transmitting and the receiver antenna is -3


300
1 meter. After all the DSP programs have been set up, then
-4
the program is started. The profile condition is under the 350
surface when the program is running and 1D plot 0 2 4 6 8 10
position (meter)
consecutive signal amplitude shown in Fig. 6 and Fig. 7. Fig. 9. GPR data image after dewowing.
In Fig. 6 and Fig. 7, it can be seen that GPR scans a
surface profile where the condition of the GPR antenna does 4
x 10
4 1D Plot Line 0 Trace 1 Dewow Data

not exactly touch the ground, so it appears in the image an


3
area where EM signals from GPR propagate in the air before
entering the ground. 2
signal amplitude (mV)

B. Data Filtering Process 1

Dewow Process and Time Zero Correction: In Fig. 6 0

above, there is no known information related to the -1

unprocessed survey GPR data. Therefore, processing is -2

extremely needed for the GPR data. In processing GPR data,


the first thing to do is to return the signal to the actual -3

position, because there is a pause time when the data is -4

released at the time of data acquisition on the field. The -5


0 100 200 300 400 500 600 700 800 900 1000
time-zero correction process is performed to eliminate the two way travel time (ns)
signal pauses. Fig. 10. GPR trace data image 1D plot after dewowing.

Line 0 4 x 10
4 1D Plot Line 0 Trace 1 (Before and After Dewow)
x 10 4
0 before dewow
3
after dewow
3

50
2 2
signal amplitude (mV)
two way travel time (ns)

100 1

0
150
0
-1

200
-2
-1

250 -3

-2
-4
300
-5
-3 0 100 200 300 400 500 600 700 800 900 1000

350 two way travetime (ns)


0 2 4 6 8 10
position (meter) Fig. 11. Comparison of trace data images ratio of 1D plot before and after
dewowing.
Fig. 8. GPR data image after time zero correction.

Fig. 8 presents the profile of GPR data after time zero C. Increasing Gain on Data
correction. In this time zero correction process, the Radar waves that propagating through the ground will
calculation of the frequency, wave propagation time getting muffled quickly. It means that the deeper the radar or
between the transmitter antenna and receiver antenna, and electromagnetic wave propagates, the weaker signal power
then eliminates the amplitude value of the signal which is which causes us can not "view" more when the signal
close to zero and interpolates for each GPR trace data. bounces below the surface and arrives at the surface. One of
After time zero correction is successfully done, the next the ways to make the signal which has propagated deeper

771
International Journal of Machine Learning and Computing, Vol. 9, No. 6, December 2019

can be seen or read is strengthened them by multiplying the a difference at the bottom because the signal on the lower
signal with a certain amplifier factor. This called as gain the profile is stronger. Strengthening the amplitude signal at the
data. There are two methods that used in this study, the first bottom of the profile and attenuating the signal amplitude at
method is to increase the power of signal data with a large the top of the profile is shown in Fig. 13. As seen in Fig. 13,
factor for signals that originating from the deeper subsurface the more located under the profile, the bigger signal strength
and multiplying it with a smaller factor for signals coming can be gained.
from the subsurface that are not too deep. This method is Automated gain control (AGC): The profile result of GPR
called time power amplitude gain correction. With this first data processing by increasing gain data using the automated
method, the GPR data image on the top layer will be a little gain control method can be seen in Fig. 14. The way to do
unclear, but the bottom layer will be strong enough. The that all is by setting up a time (depth) window and the power
other method is to multiply all signals from all layers at in each window to be the same. Therefore, we highly need
various depths with the same specific factors. This method to assign the window width. In each window, will be given
called as automated gain control. The choice of this the same strengthen for all. Fig. 15 shows the same
strengthening data signal power method depends on the type strengthen for each window.
of data or target that wanted to. Line 0
Time-power gain (TPOW): In the time power amplitude 0

gain correction method, the signal is amplified along the 50


0.8

axis of the two-way propagation time and multiplies each 0.6

trace with the power of t (tα). If we assume α = 1, then we

two way travel time (ns)


100 0.4

increase signal power linearly at each propagation time. If 150


0.2

we assume t = 2, then the power will increase quadratically 0

throughout the propagation time. In this picture below, we 200


-0.2

use α = 2. 250 -0.4

-0.6

Line 0 7
300
x 10
0 -0.8

2
350
50 0 2 4 6 8 10
1.5 position (meter)
two way travel time (ns)

100 1 Fig. 14. Image profile of GPR data after being strengthened by the
automated gain control method.
150 0.5

0 1D Plot Line 0
200 1

-0.5
0.8

250 -1
0.6

300 -1.5
0.4
signal amlitude (mV)

-2
0.2
350
0 2 4 6 8 10
0
position (meter)

Fig. 12. Image profile of GPR data after being strengthened by the time -0.2

power gain method. -0.4

-0.6
x 10
7 1D Plot Line 0
2

-0.8

1.5
-1
0 100 200 300 400 500 600 700 800 900 1000

1
two way travel time (ns)
signal amplitude (mV)

0.5
Fig. 15. 1D plot for signal trace after being strengthened by the automated
gain control method
0

Line 0
0
-0.5

1
-1
2

-1.5
3
Depth (meter)

4
-2
0 100 200 300 400 500 600 700 800 900 1000
5
two way travel time (ns)
6
Fig. 13. 1D plot for signal trace after being strengthened by the time power
7
gain method.
8

Fig. 12 presents an image that similar to the image after 9

the dewowing process. The difference can be seen on the 10

deeper surface. In the picture with the time power method, 0 2 4 6 8 10


position (meter)
the profile data on the upper surface does not change too Fig. 16. Profile of GPR data images with two-way travel time converted to
much, but the signal conditions only weaken, while there is depth with velocity = 0.06 m / ns.

772
International Journal of Machine Learning and Computing, Vol. 9, No. 6, December 2019

The profile pictures that have been described before, all techniques for criminal gravesites location,” in Proc. 2016 16th
International Conference on Ground Penetrating Radar (GPR), 2016,
images are presented in two-way travel time. This two-way pp. 1–5.
travel time can be converted to a depth. To convert the two- [9] S. M. Shrestha, I. Arai, T. Miwa, and Y. Tomizawa, “Signal
way travel time to depth, the speed of the radar signal in the processing of ground penetrating radar using super resolution
technique,” in Proc. the 2001 IEEE Radar Conference (Cat.
ground needs to be known first. For the speed of
No.01CH37200), pp. 300–305.
electromagnetic waves in the media can be seen in Table I. [10] W. F. Lang and M. X. Lin, “A practical ground penetrating radar
The speed of propagation used in GPR data is 0.6 m/s. Fig. signal processing method based on wavelet transform,” in Proc. 2010
16 showing the GPR data profile where two-way travel time Second International Conference on Communication Systems,
Networks and Applications, 2010, pp. 452–455.
is converted to depth. [11] X. Xiao, A. Ihamouten, G. Villain, and X. Derobert, “Parametric
study on processing GPR signals to get a dispersion curve,” in Proc.
the 15th International Conference on Ground Penetrating Radar,
2014, pp. 575–580.
V. CONCLUSION AND FUTURE WORKS [12] Y. Shkvyrya, “Algorithm of data processing in GPR,” in Proc. 2010
The goal of this study is to provide a free program to International Kharkov Symposium on Physics and Engineering of
Microwaves, Millimeter and Submillimeter Waves, 2010, pp. 1–1.
process and visualize the GPR data. This program is still [13] R. N. Nagashree, N. Aswini, A. Dyana, and C. H. S. Rao, “Detection
under development. This program in this study follow the and classification of ground penetrating radar image using textrual
phase to processing GPR data as in [3], start with preparing features,” in Proc. 2014 International Conference on Advances in
Electronics Computers and Communications, 2014, pp. 1–5.
the data, dewow, time zero correction, filtering, [14] [14] C. Schmelzbach and E. Huber, “Efficient deconvolution of
deconvolution, velocity analysis, elevation correction, ground-penetrating radar data,” IEEE Trans. Geosci. Remote Sens.,
migration and depth conversion. And the result of this phase vol. 53, no. 9, pp. 5209–5217, Sep. 2015.
can be seen at this paper.
Next step of this study is to convert it into freeware Baso Maruddani was born in Makassar,
program so that this program can use to many other people. Indonesia. He completed his bachelor’s degree in
2005, master’s degree in 2007, and doctorate
Our plan is to convert it into Octave Program.
degree in 2013, all are from Institut Teknologi
Bandung. Maruddani has all degrees in electronics
ACKNOWLEDGMENT and telecommunication fields, especially in
wireless communication and signal propagation.
The authors wish to thank to Head of Institute of Since 2008 to the present, he is a lecturer and
Research and Community Service, Universitas Negeri researcher at Universitas Negeri Jakarta (UNJ).
Jakarta and Head of DJA Institute which funding this Aside from being a lecturer and researcher, he also
became a consultant in the telecommunications sector. He once worked at
research. Aalborg University as a researcher in the Erasmus Mundus's Mobility for
Life program during 2010-2011. Dr. Baso Maruddani is also the head of
REFERENCES DJA Institute, a foundation which engaged in education, health and ICT
applications in the fields of education and health.
[1] K. Budiono, Handoko, and U. Hermawan, “Penafsiran struktur
geologi bawah permukaan di kawasan semburan lumpur sidoarjo
berdasarkan penampang Ground Penetrating Radar (GPR),” J. Geol. Efri Sandi was born in Rumbai, Indonesia, he
Indones., vol. 5, pp. 187–195, 2010. obtained his bachelor’s degree at Universitas
[2] E. Huber and G. Hans, “RGPR — An open-source package to process Negeri Padang in 1999, his master’s degree at
and visualize GPR data,” in Proc. 2018 17th International Conference Universitas Trisakti in 2004 and his doctorate
on Ground Penetrating Radar (GPR), 2018, pp. 1–4. degree at Universitas Indonesia in 2017. All of
[3] H. M. Jol, Ground Penetrating Radar Theory and Applications, them are majoring in electronics and
Oxford: Elsevier Science, 2008. telecommunication, especially in antenna and
[4] Y. Erdogan, Parametric Study and Design of Vivaldi Antennas and propagation. Since 2008 to the present, he is a
Arrays, Ankara, Turki: Middle East Technical University, 2009. lecturer and researcher at Universitas Negeri
[5] P. J. Gibson, “The vivaldi aerial,” in Proc. 1979 9th European Jakarta (UNJ). Dr. Efri Sandi also became a
Microwave Conference, 1979, pp. 101–105. consultant in the telecommunication sector. Many projects related
[6] T. Hariyadi and M. Mukhidin, “Studi Parametrik Antena Vivaldi Slot telecommunication field are handled by him.
dengan Pencatuan Mikrostrip,” in Proc. Seminar Nasional Teknologi
(SENATEK), 2015, pp. 397–403.
[7] R. Rajaraman, Design of a Wideband Vivaldi Antenna Array for the
Snow Radar, Kansas: University of Kansas, 1999.
[8] N. Pudova, A. Urusova, M. Shirobokov, and A. Marchkov,
“Developing GPR surveys, data processing and interpretation

773
Neural Network Algorithm and Application

You might also like