Remotesensing 12 01228 v2 PDF

remote sensing
Article
Application of Kinematic GPR-TPS Model with High
3D Georeference Accuracy for Underground Utility
Infrastructure Mapping: A Case Study from Urban
Sites in Celje, Slovenia
Nikolaj Šarlah 1, * , Tomaž Podobnikar 2 , Tomaž Ambrožič 2 and Branko Mušič 3
1 School Centre Celje, Higher Vocational College, Pot na Lavo 22, SI-3000 Celje, Slovenia
2 Faculty of Civil and Geodetic Engineering, The University of Ljubljana, Jamova 2, SI-1000 Ljubljana,
Slovenia; tomaz.podobnikar@fgg.uni-lj.si (T.P.); tomaz.ambrozic@fgg.uni-lj.si (T.A.)
3 Faculty of Arts, The University of Ljubljana, Aškerčeva 2, SI-1000 Ljubljana, Slovenia;
branko.music@ff.uni-lj.si
* Correspondence: niko.sarlah@gov.si; Tel.: +386-40-473-753

Received: 17 February 2020; Accepted: 10 April 2020; Published: 11 April 2020
Abstract: This paper describes in detail the applicability of the developed ground-penetrating radar
(GPR) model with a kinematic GPR and self-tracking (robotic) terrestrial positioning system (TPS)
surveying setup (GPR-TPS model) for the acquisition, processing and visualisation of underground
utility infrastructure (UUI) in a real urban environment. The integration of GPR with TPS can
significantly improve the accuracy of UUI positioning in a real urban environment by means of
efficient control of GPR trajectories. Two areas in the urban part of Celje in Slovenia were chosen.
The accuracy of the kinematic GPR-TPS model was analysed by comparing the three-dimensional
(3D) position of UUI given as reference values (true 3D position) from the officially consolidated
cadastre of utility infrastructure in the Republic of Slovenia and those obtained by the GPR-TPS
method. To determine the reference 3D position of the GPR antenna and UUI, the same positional
and height geodetic network was used. Small unmanned aerial vehicles (UAV) were used for
recording to provide a better spatial display of the results of UUI obtained with the GPR-TPS method.
As demonstrated by the results, the kinematic GPR-TPS model for data acquisition can achieve an
accuracy of fewer than 15 centimetres in a real urban environment.
Keywords: kinematic GPR-TPS model; self-tracking terrestrial positioning system; underground

utility infrastructure; unmanned aerial vehicle; horizontal accuracy; vertical accuracy; real
urban environment
1. Introduction
Ground-penetrating radar (GPR) is one of the most popular subsurface geophysical methods
adopted for acquisition, position and mapping of underground utility infrastructure (UUI). The basics
of the GPR principle are already well explained [1–3]. The unseen network of UUI is very complex in
any urban environment [4]. Their importance and utility infrastructure cadastre would not be obvious
until hazards and problems arise, such as a gas explosion, road collapse due to subsurface wash-out,
water leakage and seepage to the road surface, etc [4–7].
The consolidated cadastre of utility infrastructure (CCUI) in the Republic of Slovenia was designed
in 2004 as a centralised point of the utility infrastructures (UI) owners, who supply the system with
data, and data users. Its purpose is to register all the UI, especially for local and state spatial planning,
to prepare the database for the registration of legal rights (ownership) on UI, and to establish a system
Remote Sens. 2020, 12, 1228; doi:10.3390/rs12081228 www.mdpi.com/journal/remotesensing

Remote Sens. 2020, 12, 1228 2 of 26
for the prevention of damage [5]. The following types of UI are registered: energy supply infrastructure
(electric energy, natural gas, heating, oil transport), water-distribution system, stormwater drainage,
sewer system, street lighting and traffic lighting cables and electronic communications networks [8,9].
The Surveying and Mapping Authority of the Republic of Slovenia (SMA) manages the entire system,
while owners of UI are responsible for providing data, and geodetic surveyors are responsible for
horizontal and vertical positions. The allowable horizontal and vertical position errors are guided
by the ASCE 38-02 [10] standard from the USA, AS 5488-2013 [11] from Australia, and ICE PAS
128:2014 [12] from the UK. These standards categorise the utility survey results into four quality levels
(QL): QL-A, QL-B, QL-C and QL-D. QL-A gives the highest accuracy while QL-D gives the lowest.
These standards categorise the GPR method as the second-best quality level (QL-B). The horizontal
and vertical accuracy of registered UI in CCUI are introduced by exchange format and the code lists of
CCUI [13] and the guidelines for underground utility mapping in Slovenia [14] (see Table 1). The QL-B
is carried out by horizontal accuracy of the coordinates of points, assessed with a semi-major axis of a
standard ellipse of less than 40 cm. Furthermore, the vertical accuracy of UI estimated by the standard
deviation of the height must not be less than 50 cm (Supplementary Materials).
Table 1. Horizontal and vertical accuracy requirements in Slovenia (global navigation satellite system
(GNSS); terrestrial positioning system (TPS) surveying setup.
Accuracy
Quality Level SUB-QL Survey Method
Horizontal Vertical
[cm] [cm]
Desktop study of available records
QL-D and interviews with the local people; / /
other methods
Non-geodetic surveying (mobile
more than
QL-C3 GNSS receivers); old analogue ±1000
±100
QL-C geodetic plan 1:2880 and 1:5000
Geodetic surveying (TPS or GNSS) of
ground features (valves, manhole,
QL-C2 ±500 ±100
hydrant, transformer room, etc); old
analogue geodetic plan 1:1000
Geodetic surveying (TPS or GNSS
methods) immediately after filling up
QL-C1 ±100 ±100
the trenches; old analogue geodetic
plan 1:500
Geophysical methods (pipe cable
QL-B locator, low-frequency ±40 ±50
electromagnetic methods and/or GPR)
Open-up geodetic surveying (TPS or
QL-A ±10 ±10
GNSS) where the utility is exposed
The key motivation for the research was to demonstrate the usefulness of the kinematic GPR
model for the acquisition, processing and visualisation of UUI in a real urban environment at least with
QL-B. As demonstrated by the results in Šarlah et al. [5], the developed GPR model with a kinematic
GPR and self-tracking (robotic) terrestrial positioning system (TPS) surveying setup (GPR-TPS model)
for data acquisition is capable of achieving an accuracy of less than 10 centimetres in real urban testing
pools. The GPR-TPS model was created for high-accuracy underground utilities Three-dimensional
(3D) mapping in real-time and only minor adaptations would be necessary for the transfer to fields
of GPR investigation in real urban environments. Three-dimensional scan continuous reflections
resulted from hyperbolas from a series of parallel B-scans that can be mapped clearly and defined as
UUI. By contrast, in a single 2D B-scan traverse, any hyperbolic reflection can be either a utility or
some other anomalies with significant dielectric contrast to the host soil, such as boulders [4]. A 3D
Remote Sens. 2020, 12, 1228 3 of 26
scan was undertaken conventionally by traversing a GPR antenna in an X-Y orthogonal grid on the
ground [6,8,15,16]. To eliminate the use of rectangular grids for positioning of a GPR antenna, a global
navigation satellite system (GNSS) and TPS can be used by mounting a GNSS receiver or a 360◦
reflector (prism) on the centre top of the GPR antenna [5,17–22]. Hence, accurate GNSS positioning is
not always available in urban GPR surveys and an alternative method is required.
Table 2 reports the previous efforts spent on how UUI is positioned and mapped, and how their
conditions can be assessed by GPR. For many types of research, the horizontal and vertical accuracy
of UUI achieved by GPR is not known due to the unknown true position of UUI. Elucidations of the
determined UUI are concealed by the effect of the various surface noises, the positioning method of
GPR, the location of test field or urban site, the dimensions and electrical properties of the subsurface
and UUI, and the different bandwidth applied based on the depth ranges.
Table 2. Some examples of applications of ground-penetrating radar on underground utility

infrastructure detection.
Positioning of Achieved
References Year Bandwidth [MHz] Test Field the GPR Accuracy
Antenna Horiz./Vertic.
Ayala-Cabrera et al. [6] 2011 1500 Test field Orth. grid Unknown
163
Urban site I Profiles cm/Unknown
Bilal et al. [23] 2018 Unknown
Urban site II Profiles 100
cm/Unknown
Dou et al. [24] 2016 Unknown Urban site Profiles 30 cm/30 cm
Urban site I Profiles 30 cm/30 cm
Chen and Cohn [25] 2011 Unknown
Urban site II Profiles 40 cm/40 cm
Cheng et al. [26] 2013 100, 270 and 400 Urban site Orth. grid 33 cm/61 cm
Test
Gabryś et al. [27] 2019 250, 500 and 700 TPS and GNSS 15 cm/Depth
field/Urban site
Ghozzi et al. [28] 2018 400 Urban site Orth. grid Unknown
Grandjean et al. [15] 2000 300, 500 and 900 Test field Orth. grid Unknown
Ismail et al. [29] 2013 250 Urban site Profiles Unknown
Jaw and Hashim [18] 2013 250 and 400 Test field GNSS 10 cm/10 cm
Jeng and Chen [30] 2012 200 and 800 Urban site Profiles Unknown
Li et al. [19] 2015 800 Test field GNSS 10 cm/30 cm
Metwaly [31] 2015 400 Urban site Profiles Unknown
Mušič et al. [16] 2011 400 Urban site Orth. grid Unknown
Porsani et al. [32] 2012 200 Urban site Profiles Unknown
300, 500, 800, 900
Sagnard et al. [33] 2016 Test field Profiles Unknown
and 1500
Šarlah et al. [5] 2019 270, 400 and 900 Test field TPS 8 cm/12 cm
Note: Profiles: A few surface parallel GPR profiles located above the underground utility infrastructure (UUI);
Unknown: Accuracy has not been estimated; Depth: Depends on the depth (10% of the depth of the UUI).
As is seen from Table 2 some studies were focused on determining the positional and height
accuracy of UUI in a test environment [5,18,19,27] and a few in a real urban site [23–26]. None of them,
other than Šarlah et al. [5] and Gabryś et al. [27], use TPS to determine the position of the GPR antenna
in kinematic mode. In Dou et al. [24] and Bilal et al. [23], a unique marching cross-section algorithm and
Bayesian mapping model with implementing various machine-learning techniques for automatically
locating UUI segments by fusing data from multiple sensors are introduced. All profiles measured
with a GPR and other sensors were later calibrated on a known previously established coordinate
system determinate with TPS. Despite the actual non-linear movement, strict linear movement is
Remote Sens. 2020, 12, 1228 4 of 26
predisposed from the starting to the final point of the profile, which significantly reduces positional
accuracy. In [25], 12 trial pits were excavated to obtain the true position and height of UUI in the
test site. The true position and heights of the UUI between the pits is assumed rather than measured.
The measuring methods and the accuracy of true position and heights of UUI are not cited. By using
GPR data, 60.7% of the UUI in the urban site were detected closer than 30 cm according to the assumed
true position and heights. In [23] the positional accuracy given with the mean spatial error S(x;y) is
163 cm for urban site I and 100 cm for urban site II. S(x;y) represents the mean spatial distance from
the true to the estimated position. In [25], the individual hypothesised detections from a group of
parallel GPR scans are used to determine the approximate location and direction of a UUI segment.
The Bayesian data fusion algorithm is proposed for automatic UUI mapping. Determining the actual
truth position and heights of UUI would involve excavation. The measuring methods and the accuracy
of true position and heights of UUI are not cited. The positional and heights accuracy given with the
mean spatial error E(x;θ) is 30 cm for real site I and 40 cm for real site II. E(x;θ) represents the mean
spatial distance from the true to the estimated 3D position. Cheng et al. [26] cite the measurements
with 270 MHz and 400 MHz antennas with the calibrated measuring wheel on a completely horizontal
plain relief of three chosen real test urban sites. All distances measured with a GPR measuring wheel
were later calibrated on known distances of the previously established orthogonal grid. The measuring
methods and the accuracy of true position of UUI and orthogonal grid are not cited. Radargrams are
processed with a zero-time correction, infinite and finite impulse response filtering, deconvolution
and migration. The method to determine the estimation of velocity of electromagnetic wave (EMW)
propagation and the used method of selecting the objects in the medium is not cited. The average
positional accuracy of 400 MHz antenna of three detected pipes, given with the positional (radial) root
mean square error (RMSEpos ), is 33.5 cm, while the average vertical accuracy of three detected pipes
and one electric cable is 61.3 cm, given the height RMSE.
2. Description of Sites
Two areas in the urban part of the Municipality of Celje in Slovenia were chosen, where the
GPR-TPS model [5] was applied. A few basic criteria led towards selection of the representative testing
area: appropriate relief; appropriate content of the area (coating with asphalt mixtures); the surface is a
road or a pavement; there should be at least one UUI in the area, the position of which was determined
with geodetic methods before the trenches were refilled; recorded in the CCUI; the existence of a
reference basis or a geodetic network used when the geodetic surveying was done; an absence of EMW
sources which can cause disturbances in GPR measurements due to interference (e.g., power lines,
transformer stations, radio and television transmitters, GSM network transmitters); and an absence of
the local deformities of the earth’s surface (areas with noticeable subsidence, avalanche and erosive
areas). UUI positions were given in a national coordinate system and UUI height as the above sea-level
altitude, given in the system of normal orthometric heights.
The real urban site I is situated west from the Hudinja watercourse and south of Bežigrajska street
in Celje. All new UUI (sewers, water pipes, gas pipes and power lines) were newly built in the area.
The intersection or the juncture of the extension of the Dečkova and Njegoševa streets in the Lava city
quarter in Celje was selected as the real urban site II (see Figure 1). The frequency of traffic was the
cause of the impediment, since it did not allow execution of the measurements on the entire area of the
intersection. The main reason for the selection of the area was the reconstruction or renovation of the
water and gas distributional supply system at the beginning of June 2012.
Remote Sens. 2020, 12, 1228 5 of 26
Remote Sens. 2020, 12, x FOR PEER REVIEW 5 of 25
b) a) c)
Figure 1.
Figure 1. (a) On the
the map
map the
the position
positionofofreal
realurban
urbansites
sitesI and
I andII II
areare
marked;
marked; (b)(b)Presentation
Presentationof aof a
total of
total of 16,878
16,878 points
points determined
determinedusing usingthe
theground-penetrating
ground-penetrating radar with
radar a terrestrial
with a terrestrialpositioning
positioning
system (GPR-TPS)
system (GPR-TPS) modelmodel(red
(redpoints
pointsshow
show 28 28
profiles in the
profiles S–NS–N
in the direction) i.e. real
direction) i.e.,urban site II; site
real urban (c) II;
Presentation of a total of 15,423 points determined using the GPR-TPS model
(c) Presentation of a total of 15,423 points determined using the GPR-TPS model (red points show (red points show 19 19
profiles in the S–N direction) i.e. real urban site
profiles in the S–N direction) i.e., real urban site I. I.
The structure
Prior and of
to the filling thickness of thealayer
the trenches, of terrestrial
classic the pavement and surfaces
geodetic are summarised
measurement of the UUI in thethe
from
project documentation (see Table 3).
poligonometric network of the I. order was executed. The accuracy of the true position and height of
UUI determined using the geodetic methods is unknown. An identical poligonometric network as
Table 3. Thickness of real urban sites layers.
well as the starting and orientation points were used as a geodetic basis to determine the position of
the GPR antenna with the GPR-TPS model.site I
Real urban Real urban site II
The structure and thickness of the layer of the pavement and surfaces are summarised in the
project documentation (see Table Material
3). Depth Material Depth
[cm] [cm]
Surface Course bituminous Thickness of real4urban sites layers.
Table 3.concrete asphalt concrete 4
bituminous base 9 asphalt base 8
Real Urban Site I Real Urban Site II
Base Course / / crushed rock aggregate 32
Material Depth [cm] Material
(limestone grains 0–32 mm) Depth [cm]
Surface Course
Subbase Course bituminous
crushed rock aggregate 4 asphalt concrete
45
crushed rock aggregate 404
concrete
(limestone grains 0–125 mm) (limestone grains 0–125 mm)
bituminous base 9 asphalt base 8
Subgrade
Base Course /soil with rock /
/ crushedsoil
rockwith rock
aggregate / 32
(limestone grains 0–32 mm)
Real urban
Subbase Coursesite I: Technical
crushed rockdocumentation
45 cites the gas pipes,rock
crushed industrial
aggregatewater supply40pipes,
power cable and meteor,aggregate
technological and faecal sewage (limestone
pipes ingrains 0–125 mm)
the chosen area. All UUI lay on
the 10 cm thick sand (limestone
base and grains
is covered with sand up to a height of 15 cm above the top (apex)
0–125 mm)
point. The
Subgrade
data about voltage, depths
soil with rock
and characteristics
/
of power lines are unknown (see Table
soil with rock /
4).
Real urban site II: Water pipes, represent the reconstructed water system of the addressed area.
The data show the presence pipes of industrial plumbing, which was built in 1991. The notes
Realthe
indicate urban site I:ofTechnical
existence documentation
a signal cable duct for thecites
needsthe
ofgas pipes, of
telemetry industrial water supply
the waterworks; pipes,
its spatial
power
position is unknown. The material, dimensions, position and height of meteor sewage data are alsothe
cable and meteor, technological and faecal sewage pipes in the chosen area. All UUI lay on
10 cm thick(see
unknown sandTable
base and
4). Theis covered with and
gas supply sandthermal
up to a height
energy of 15 cm above
company the top
in charge of (apex) point.
the area
The data about
discussed cites voltage,
the data depths
about gasand characteristics
pipes, and two hotofwater
powerpipes
lineswith
are unknown (see of
a temperature Table
90/70 4).°C, of
the culvert (1.70 m × 0.80 m). The data about the dimensions of the culvert and the thickness of the
isolation (23 mm) of heating pipes are summarised from the geodetic record of the hot water system,
from 1988. The thickness of the isolation at a later renovation is unknown. The data about the
position and the height of the industrial water supply and heating system were in the past obtained
based on the cadastre of communal infrastructure. The data obtained through digitalisation of the
plans of the cadastre of communal infrastructure in the 1:1000 scale cannot act as a referential basis
for assessing the accuracy of the used GPR-TPS model, however, it can be an appropriate basis. The
original tachymetric records, from which the true position and height could be calculated, were not
preserved.
Remote Sens. 2020, 12, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing

Remote Sens. 2020, 12, 1228 6 of 26
Table 4. A summary of all the buried pipes, electrical cable, culvert, electronic communication cable
duct in the real urban sites (ND—nominal diameter, OD—outside diameter, PVC—polyvinyl chloride,
PE—polyethylene, CO—concrete, DI—ductile iron, DCI—ductile cast iron, GRP—glass reinforced
polyester).
Object Nominal Diam. Real Urban

Type Material Depth [cm] Position
No. ND/OD [mm] Site [No.]
1 Gas PE 110 170 True I
2 Water DI 100 200 True I
3 Faecal sewage PVC 400 200–350 True I
4 Technological sewage PVC 400 200–350 True I
5 Meteor sewage PVC 300 130–150 True I
6 Electrical cables PVC 31 80 Unknown I
7 Gas PE 63 80 True II
8 Water DCI 200 80–100 True II
Industrial water
9 GRP 400 200 CCUI II
plumbingWater
10 Water DI 250 140 CCUI II
11 Water DI 250 140 CCUI II
12 Heating – culvert CO 170 × 80 110 CCUI II
13 Meteor sewage / / / Unknown II
14 Electronic cable duct PVC 110 80 Unknown II
15 Signal cable duct PVC 110 / Unknown II
* Consolidated cadastre of utility infrastructure (CCUI): data obtained through digitalisation of the plans at a scale
of 1:1000.
Real urban site II: Water pipes, represent the reconstructed water system of the addressed area.
The data show the presence pipes of industrial plumbing, which was built in 1991. The notes indicate
the existence of a signal cable duct for the needs of telemetry of the waterworks; its spatial position is
unknown. The material, dimensions, position and height of meteor sewage data are also unknown
(see Table 4). The gas supply and thermal energy company in charge of the area discussed cites
the data about gas pipes, and two hot water pipes with a temperature of 90/70 ◦ C, of the culvert
(1.70 m × 0.80 m). The data about the dimensions of the culvert and the thickness of the isolation
(23 mm) of heating pipes are summarised from the geodetic record of the hot water system, from
1988. The thickness of the isolation at a later renovation is unknown. The data about the position
and the height of the industrial water supply and heating system were in the past obtained based on
the cadastre of communal infrastructure. The data obtained through digitalisation of the plans of the
cadastre of communal infrastructure in the 1:1000 scale cannot act as a referential basis for assessing the
accuracy of the used GPR-TPS model, however, it can be an appropriate basis. The original tachymetric
records, from which the true position and height could be calculated, were not preserved.
3. Methods and Instrumentation

The identical system of equipment and methods to execute geophysical and geodetic measurements
as in the testing environment is used as developed in Šarlah et al. [5] in real urban sites. The geophysical
survey system GSSI SIR3000 with the holder and the nozzle for the reflector set exactly above the
central point of the GPR antenna was used. A linearly polarised dipole-shielded antenna with a central
frequency of 400 MHz was used. The domain of triggering the GPR signal in the dependency of the
length was enabled with the corresponding measuring wheel. Acquisition settings for the subsurface
profiles were set to 50 scans per metre, 1024 samples per scan and 16 bits per sample, with a range equal
to 75ns. The system upgrade with the interface (Bluetooth serial port RS 232) with an outer dipole
antenna (BT-232B-E) enabled the establishment of a Bluetooth connection between the TPS and GPR
used in real-time. The frequency of the terrestrial kinematic measurement was 4–5 Hz, while the speed
of the moving reflector was ≈ 0.5m/s (see Figure 2). While presupposing the application of the same
system of equipment when the measurements were executed, the values of the latency—calculated
on the base of a real urban testing pool—were obtained. Additionally, the suggested model for the
Remote Sens. 2020, 12, 1228 7 of 26

processing of the radargrams, developed in Šarlah et al. [5], was obtained. The quantitative data for
wavelength.
the 3D display In of
theUUI
caseswasof real urbanthrough
obtained site I andthe
II, selected
the yields were a minimum
collection PS// of 0.98for
of the procedures m processing
and PS
⊥ = 1.47 m with 400 MHz (the estimated minimal velocity in sites is 0.087 m/ns; minimal wavelength
the radar sections and an analysis of individual radar reflections [34,35], which are already fulfilled
is 0.245
with them). This further
GPR-TPS model. confirms that PSwere
The profiles = 0.5acquired
m was the inappropriate choice. e.g., using a meandering
different directions,
The Leica TCRP 1201+ electronic tachymeter, along with the
data acquisition, the corresponding profiles were flipped in distance direction automatic target using
recognition
the flip(ATR)
profile
system, enables automatic target tracking that allows automatic angle and length measurements of
function. The distance/space between the profiles (PS) was ≈ 0.5 m. In addition, the angle between
a reflector. The functionality of the ATR significantly diminishes operator errors; however, it is also
profiles and a linear object has a significant impact on GPR response. In the best case, GPR profiles
susceptible to rough errors when tracking a reflector in the same direction [37]. The declared
are perpendicular to a linear object (PS⊥ ) and a larger profile spacing is acceptable. In terms of the
standard deviation of the measured angles in the ATR mode, according to the ISO standard 17123-3,
worst case, the profiles are parallel with a linear object (PS// ), which means that the response of the
is σα = 1", the standard deviation of the measured lengths with a standard reflector is σd ≤ (1 mm; 1.5
GPR will not be obvious enough and a smaller profile spacing is required. In sum [36], a suggestion
ppm) [38]. The Leica GRZ4 360° reflector was used at terrestrial kinematic observations. For the
for the selection of a suitable profile spacing is PS// ≤ 4λ and PS⊥ ≤ 6λ; where λ is the GPR wavelength.
Leica GRZ4 360° reflector the manufacturer states the errors in horizontal and vertical directions are
In the cases
several of real [37].
millimetres urban Thesite I and II, the yields
dual-frequency Topconwere a minimum
HiPer PS// of 0.98
Pro GNSS receiver wasm andfor
used ⊥ =real-
PSthe 1.47 m
with
time kinematic method of GNSS measurements. Figure 2 shows the upgraded SIR 3000 GPR system m).
400 MHz (the estimated minimal velocity in sites is 0.087 m/ns; minimal wavelength is 0.245
This TPS. confirms that PS = 0.5 m was the appropriate choice.
withfurther
a) b) c)
Figure 2. (a)
(a) Leica
Leica TCRP
TCRP1201+
1201+electronic
electronictracking
trackingtotal
totalstation with
station automatic
with automaticrecognition of the
recognition of the
reflector located in testing field I; (b) SIR-3000 GPR system, 400 MHz mono-static antenna:
reflector located in testing field I; (b) SIR-3000 GPR system, 400 MHz mono-static antenna: (1) tow bar, 1) tow
bar,measuring
(2) 2) measuring wheel,
wheel, 3) Leica
(3) Leica GRZ4GRZ4 reflector
reflector andand
(4)4)adapter
adaptermade
madefor
forreflector;
reflector;(c)
(c)control
controlunit
unitand
and Bluetooth RS 232 adapter (BT-232B-E): (1) external dipole
Bluetooth RS 232 adapter (BT-232B-E): (1) external dipole antenna. antenna.
Photogrammetric
The Leica TCRP 1201+recording of the tachymeter,
electronic real urban site II with
along with small unmannedtarget
the automatic aerial recognition
vehicles (UAV)
(ATR)
was performed
system, enablesfor better presentation
automatic of the results
target tracking using automatic
that allows the GPR-TPS model
angle andoflength
the certain UUI. In
measurements
remote-sensing
of a reflector. The applications UAVofare
functionality themostly used for photogrammetric
ATR significantly purposes,
diminishes operator the production
errors; however, it is
of 3D-models
also susceptible oftoobjects,
rough digital surfacetracking
errors when models,a digital
reflectorterrain
in themodels (DTM) and
same direction [37].orthophotos
The declared
[39,40]. The Mikrokopter Hexa XL UAV system with six rotors, three of which
standard deviation of the measured angles in the ATR mode, according to the ISO standard rotate in a clockwise
17123-3,
is σα = 1”, the standard deviation of the measured lengths with a standard reflector is σdwith
direction and three in a counterclockwise direction, was used. The UAV system is equipped ≤ (1 amm;
SONY
1.5 ppm) α7R digital
[38]. The camera system
Leica GRZ4 360which is attached
◦ reflector at the
was used at camera mount
terrestrial (see Figure
kinematic 3). Basic geo-
observations. For the
Leica GRZ4 360 reflector the manufacturer states the errors in horizontal and verticalreceiver,
referencing of the
◦ captured photographs enables the built-in low-cost differential GNSS directions
gyro
are sensor,millimetres
several barometric[37].
sensor,
Themagnetic compassTopcon
dual-frequency and flight control
HiPer unit station
Pro GNSS which
receiver wasisused
usedfor
to the
control the UAV. MikroKopter MK tool software was used to programme the flights. The recording
real-time kinematic method of GNSS measurements. Figure 2 shows the upgraded SIR 3000 GPR
system with TPS.
Photogrammetric recording of the real urban site II with small unmanned aerial vehicles (UAV)
was performed for better presentation of the results using the GPR-TPS model of the certain UUI. In
remote-sensing applications UAV are mostly used for photogrammetric purposes, the production of
3D-models of objects, digital surface models, digital terrain models (DTM) and orthophotos [39,40].
The Mikrokopter Hexa XL UAV system with six rotors, three of which rotate in a clockwise direction
and three in a counterclockwise direction, was used. The UAV system is equipped with a SONY α7R
digital camera system which is attached at the camera mount (see Figure 3). Basic geo-referencing of the
captured photographs enables the built-in low-cost differential GNSS receiver, gyro sensor, barometric
sensor, magnetic compass and flight a) control unit station which is used b) to control the UAV. MikroKopter
MK tool software was used to programme the flights. The recording data were processed using the
Figure 3. (a) Unmanned aerial vehicle—hexacopter; (b) ground control station.
was performed for better presentation of the results using the GPR-TPS model of the certain UUI. In
remote-sensing applications UAV are mostly used for photogrammetric purposes, the production
of 3D-models of objects, digital surface models, digital terrain models (DTM) and orthophotos
[39,40]. The Mikrokopter Hexa XL UAV system with six rotors, three of which rotate in a clockwise
direction and2020,
Remote Sens. three
12, in a counterclockwise direction, was used. The UAV system is equipped with8 of
1228 a 26
SONY α7R digital camera system which is attached at the camera mount (see Figure 3). Basic geo-
referencing of the captured photographs enables the built-in low-cost differential GNSS receiver,
Imagine UAV programme module in the Erdas Imagine 2015 environment. A detailed description of
gyro sensor, barometric sensor, magnetic compass and flight control unit station which is used to
the acquisition and processing is found in the Appendix A.
control the UAV. MikroKopter MK tool software was used to programme the flights. The recording
a) b)
Figure
Remote 3.2020,
Sens. (a) Unmanned
Figure
12, aerial
(a)
3. PEER
x FOR vehicle—
Unmanned
REVIEW hexacopter;
aerial (b) ground control
vehicle—hexacopter; station.
(b) ground control station. 8 of 25
3.1. Real
3.1. Real Urban
Urban Site
Site II
Measurements on
Measurements onnine
ninerandom
randomprofiles
profilesininthe S-N
the direction
S-N (red
direction points)
(red andand
points) 10 profiles in the
10 profiles in N-S
the
direction (grey points) were executed (see Figure 4). All profiles were recorded on a plane
N-S direction (grey points) were executed (see Figure 4). All profiles were recorded on a plane withwith an
incline
an lower
incline than
lower 0.9%.
than TheThe
0.9%. lengths of the
lengths profiles
of the were
profiles from
were 11 to
from 1122
to m.
22 m.
Figure 4. Presentation
Figure 4. Presentation of
ofaatotal
totalofof15,423
15,423points,
points,determined
determinedbyby
thethe GPR-TPS
GPR-TPS model
model (red(red points)
points) on
on the
the geodetic plan showing UUI (see Table
geodetic plan showing UUI (see Table 4). 4).
Figures 5–8 and Table 5 show show the

the data
data processing
processing using
using the
the GPR-TPS
GPR-TPS model.
model. According to the
described in Šarlah et al. [5], the third phase of the processing was adapted
data processing model described
to the demands of real test site I. The time cut function of the radargram was intentionally left out
due to the explicit display of the depth range and the unrecognisability of target objects in the lower
layers due to the presence of groundwater. The The f-k
f-k filter
filter was
was omitted
omitted from
from the model due to the
non-disturbing easy
easy air
airreflections
reflectionsofofa nearby
a nearbyobject. Process
object. of the
Process subtracting
of the average
subtracting of traces
average was
of traces
used used
was on 35ontraces since itsince
35 traces represents a great probability
it represents for elimination
a great probability of the important
for elimination of the horizontal
important
reflections. The
horizontal estimation
reflections. Theof estimation
velocity of EMW propagation
of velocity of EMW was executed without
propagation wasknowing
executedthewithout
depths
knowing the depths of target UUI in combination with the method of manual hyperbola fitting and
the layer velocity method determined using individual layers of pavement structure (see Figure 6)
[5]. Time to depth conversion was limited to a depth of 3.2 m. The topographic correction was
preserved in the model despite an extremely small incline (see Figure 7). In relation to data
processing, more details can be found in [5,41].
Remote Sens. 2020, 12, 1228 9 of 26
of target UUI in combination with the method of manual hyperbola fitting and the layer velocity
Remote Sens. 2020,
method 12, x FOR PEER
determined usingREVIEW layers of pavement structure (see Figure 6) [5]. Time to9 of
individual 25
depth
conversion was limited to a depth
Manual of 3.2
signal gain m. The
—gain topographic
factor [dB] correction was preserved
0–25 in the model
despite an extremely Manual
small signal
inclinegain—gain
(see factor
Figure 7). [dB]
In relation to data 0–25more details can be
processing,
Time to depth conversion—max depth axis [m] 3.2
found in [5,41]. Time to depth conversion—max depth axis [m] 3.2
a)
a)
b)
b)
Figure 5. Processing of profile 8 using the GPR-TPS model: (a) rough GPR observations; (b)
Figure
Figure5.5.Processing
Processing of profile8 8using
of profile using
the the GPR-TPS
GPR-TPS model:model: (a) rough
(a) rough GPR observations;
GPR observations; (b)
(b) subtracting
subtracting a constant shift on the radargram with a time interval of the calculation of an average of
subtracting a constant shift on the radargram with a time interval of the calculation
a constant shift on the radargram with a time interval of the calculation of an average of 50–70 of an average ofns;
50–70 ns; manual determination of zero time at the first positive apex point of signal using a manual
50–70 ns; manual determination of zero time at the first positive apex point of signal
manual determination of zero time at the first positive apex point of signal using a manual setting using a manual
setting of zero (4.41 ns), spatial filter of background removal, frequency band-pass filter of tapered
setting of (4.41
of zero zero (4.41 ns), spatial
ns), spatial filter filter of background
of background removal,
removal, frequency
frequency band-pass
band-pass filterfilter of tapered
of tapered cosine
cosine window (250/310/580/680) and manual gain of amplitude alongside the profile at the
cosine
windowwindow (250/310/580/680)
(250/310/580/680) and manualand manual gain of amplitude
gain of amplitude alongside the alongside
profile atthe
the profile
expected at depths
the
expected depths of point reflectors. The gain factor of the trace signal is in the range of 0–37 dB.
expected depths of point reflectors. The gain factor of the trace signal is
of point reflectors. The gain factor of the trace signal is in the range of 0–37 dB.in the range of 0–37 dB.
a) b)
a) b)
Figure 6. Estimation
Estimation of velocity of electromagnetic wave (EMW) propagation on profile 8: (a) the
Figure
Figure 6. 6.
Estimation of of velocity
velocity of of electromagnetic
electromagnetic wave
wave (EMW)
(EMW) propagation
propagation onon profile
profile 8: 8:
(a)(a)
thethe
method
method of hyperbola fitting (examples of hyperbola fitting at a speed of 0.100 m/ns) and the method
method of of hyperbola
hyperbola fitting
fitting (examplesofofhyperbola
(examples hyperbolafitting
fittingatataaspeed
speedof
of0.100
0.100m/ns)
m/ns) and
and the
the method
method of
ofvelocity
velocity determination using individual layers of pavement structure; (b) velocity field with
of velocity determination using individual layers of pavement structure; (b) velocity fieldcoloured
determination using individual layers of pavement structure; (b) velocity field with with
coloured velocity
velocityvelocity layers (0.131 m/ns0.124
layers (0.131 —blue; 0.124 m/ns—red; 0.087 m/ns—brown).
coloured layersm/ns—blue;
(0.131 m/ns—blue; m/ns—red;
0.124 m/ns0.087
—red; m/ns—brown).
0.087 m/ns—brown).

Remote Sens. 2020, 12, 1228 10 of 26
a)
a)
b)
b)
Figure7.7.Processed
Figure Processedradargram
radargramofof profile:(a)
(a) time-depthconversion
conversionand andthe
thetopographic
topographiccorrection;
correction;
Figure 7. Processed radargramprofile:
of profile: time-depth
(a) time-depth conversion and the topographic correction;
(b) Kirchoff
(b) (b)
Kirchoff migration
migration based
based on
onontwo-dimensional
two-dimensional spatial
spatial fields
fields and
and spatially unlimited manual gain
Kirchoff migration based two-dimensional spatial fields andspatially
spatiallyunlimited
unlimitedmanual
manualgain
gain of
ofthe
of thesignal
signal (factor0–22
0–22 dB)and
and topographiccorrection.
correction.
the signal(factor
(factor 0–22dB)
dB) andtopographic
topographic correction.
a)
a)
b)
b)
Figure 8.8.Manual
Figure
Figure Manual
8. Manualselection
selection oftarget
selection
of target objects
of target onradargram
objects
objects on radargram
on radargram ofprofile
of profile
of 8:8:(a)
profile (a) selecting
8: selecting
(a) with
selecting with
with hyperboles
hyperboles
hyperboles
fitting
fitting (red
fitting
(red(redcrosses) on
crosses)on
crosses) processed non-migrated
on processed non-migrated profile;
non-migratedprofile; (b)
(b)(b)
profile; transmission
transmission
transmission and
and and corrigendum
corrigendum of
of selection
corrigendum of
selection
(yellow
selection (yellow crosses)
crosses)
(yellow on
on migrated
crosses) migrated profile.
profile.profile.
on migrated
3.2.Real
3.2. RealUrban
UrbanSite
SiteIIII
The direction
The direction of
of GPR
GPR profiles
profiles was
was determined
determined based
based on on the
the technical
technical documentation
documentation or or the
the
orientationof
orientation ofthe
themajority
majorityofofUUI.
UUI.Twenty-eight
Twenty-eightparallel
parallelprofiles
profilesin inaaS–N
S–Ndirection
directionwere
weremeasured.
measured.
All profiles were recorded on a flat asphalt surface with a very small incline at
All profiles were recorded on a flat asphalt surface with a very small incline at a GPR speed ofa GPR speed of≈0.5
≈0.5
m/s.The
m/s. Thelengths
lengthsofofthe
theprofiles
profileswere
were13.5–34.5
13.5–34.5m.
m.
Figures 9–12 and Table 6 show the capture
Figures 9–12 and Table 6 show the capture and and data
data processing
processing whenwhen using
using the
the GPR-TPS
GPR-TPS
model. According to the data processing model described in Šarlah et al. [5], the
model. According to the data processing model described in Šarlah et al. [5], the third phase of third phase ofthe
the
processing was
processing was adapted
adapted toto the
the demands
demands of of real
real test
test site
site II.
II. The
The beginning
beginning of of the
the first
first phase
phase
Remote Sens. 2020, 12, 1228 11 of 26
Table 5. The procedures and parameters used in the proposed GPR-TPS model in real urban site I.
Parameters
Process
400 [MHz]
Direct-current offset (DC Shift)—interval [ns] 50–70
Time zero correction [ns] 4.73
Manual signal gain—gain factor [dB] 0–37
Band-pass frequency with tapered cosine window [MHz] 250/310/580/680
f-k filtering limited by reflections for the positive and /
negative directions [m/ns] /
Subtracting average [traces] 35
Determination of 2D velocity field—interval [m/ns] 0.087–0.131
P
Kirchhoff 2D time migration— width [No. traces] 22
3.2. Real Urban Site II

The direction of GPR profiles was determined based on the technical documentation or the
orientation of the majority of UUI. Twenty-eight parallel profiles in a S–N direction were measured.
All profiles were recorded on a flat asphalt surface with a very small incline at a GPR speed of ≈0.5 m/s.
The lengths of the profiles were 13.5–34.5 m.
Figures 9–12 and Table 6 show the capture and data processing when using the GPR-TPS model.
According to the data processing model described in Šarlah et al. [5], the third phase of the processing
was adapted to the demands of real test site II. The beginning of the first phase contained a calculation
and corrigendum of zero time (5.20 ns), the signal delay elimination (50–64 ns) and a filter of background
removal. The manual amplitude correction (0–32 dB) and the frequency band-pass filter of the tapered
cosine window were used (230/320/580/750 MHz). The processing is the base for determining the lateral
reflections of the pavement using the picking object method. The estimation of the EMW propagation
velocity using the layer pick velocity method was the basis in the pavement structure (see Figure 10),
while the method of hyperbola fitting was used in the subgrade soil. Time to depth conversion was
limited to a depth of 3.1 m. The topographic correction was executed.
Table 6. The procedures and parameters used in the proposed GPR-TPS model in real urban site II.
Parameters
Process
400 [MHz]
DC Shift—interval [ns] 50–64
Time zero correction [ns] 5.20
Band-pass frequency with tapered cosine window [MHz] 230/320/580/750
f-k filtering limited by reflections for the positive and +0.098 to +0.057
negative directions [m/ns] −0.043 to −0.072
P
Kirchhoff 2D time migration— width [No. traces] 30
directions [m/ns] –0.043 to –0.072
Kirchhoff 2D time migration—∑ width [No. traces] 30
Remote Sens. 2020, 12, 1228 12 of 26
a)
b)
Figure
Figure9.9.Processing
Processing profile
profile 88ofof400400
MHz MHz antenna
antenna withwith the proposed
the proposed GPR (a)
GPR model: model: (a) raw
raw radargram;
radargram;
(b) DC-shift on radargram with a time interval of calculation on average between 50–64 ns, 50–64
(b) DC-shift on radargram with a time interval of calculation on average between manual
ns,determination
manual determination
of zero timeof zero time
at the at positive
first the first apex
positive apex
point of point
signalofbysignal by manually
manually setting
setting zero time
zero time
(5.20 ns),(5.20 ns),filter
spatial spatial filter of background
of background removal, bandpass
removal, frequency frequencyfilter
bandpass filter ofcosine
of the tapered the tapered
window
cosine window (250/310/580/750)
(250/310/580/750) and manually gained and manually
amplitude gained
along theamplitude along the
profile of expected profile
depths of expected
of point reflectors.
depths of point reflectors. Signal gain factor
Signal gain factor follows in the range of 0–32 dB. follows in the range of 0–32 dB.
a) b) c)
Figure
Figure10.
10.(a)
(a) Partially radargramofofprofile
Partially processed radargram profile2 of
2 of
400400
MHzMHz antennae;
antennae; (b) estimation
(b) estimation of EMWof
EMW propagation
propagation velocityvelocity using
using the the
layer layer
pick pick method
velocity velocityofmethod of the structure;
the pavement pavement(c)structure; (c)
Velocity field
Velocity field with
with coloured coloured
values values layers
by individual by individual layers (0.118
(0.118 m/ns—blue; m/ns
0.105 —blue; 0.105
m/ns—red; m/ns—red; brown,
0.102 m/ns—light 0.102
m/ns—light brown, 0.098 m/ns—brown).
0.098 m/ns—brown).
a) b) c)
Figure 11. (a) Reconstruction of the distribution waterpipe of the testing field in 2012; (b) partially
processed radargram (calculation and corrigendum of zero time, subtracting a constant shift and
Figure 10. (a) Partially processed radargram of profile 2 of 400 MHz antennae; (b) estimation of
EMW propagation velocity using the layer pick velocity method of the pavement structure; (c)
Velocity field with coloured values by individual layers (0.118 m/ns—blue; 0.105 m/ns—red; 0.102
m/ns—light brown, 0.098 m/ns—brown).
Remote Sens. 2020, 12, 1228 13 of 26
a) b) c)
Figure 11. (a) Reconstruction of the distribution waterpipe of the testing field in 2012; (b) partially
processed radargram (calculation and corrigendum of zero time, subtracting a constant shift and
filter of background removal) of a 400 MHz antenna second profile with clearly visible layers of
pavement structure; (c) determination of pavement structure layers using the semi-automatic phase
follower picking method.
The procedure of signal elimination was executed prior to the manual selection of target
objects in order to eliminate the horizontal reflections, e.g., layers of the pavement structure. The f-k
filter was used in the model where for each individual profile, the parameter of the velocity range
of a fan-shapeda) form, with which the set objective b) was reached, was defined, since the c) procedure
was notFiguresatisfactory.
Figure11.11. (a) A recognition
(a) Reconstruction
Reconstruction of analysis
of the was
the distribution done using
distribution waterpipe
waterpipe of a referential
of the
the testing manual
testing field
field in selection
in2012;
2012; (b) of the
(b)partially
partially
targetprocessed
objects (see
processed Figure (calculation
radargram
radargram 12), which are
(calculationand later
and used onof
corrigendum
corrigendum migrated
of zero
zero time,radargrams.
time, aThe
subtracting
subtracting adeconvolution
constant
constant was
shiftfilter
shift and and
additionally
filter of used on
backgroundindividual
removal)radargrams.
of a 400 MHzIt turned
antennaout to
second be useful
profile in
with real urban
clearly
of background removal) of a 400 MHz antenna second profile with clearly visible layers of pavement site
visible II in
layers the
of
area of iron
structure;heating
pavement(c) pipes,
structure; which cause
(c) determination
determination of pavement visibly multiplying
of pavement reflections.
structure
structure layers usinglayers Pavement structure
using the semi-automatic
the semi-automatic layers
phase
phase follower
estimations
follower
picking were made
picking
method. using a ‘semi-automatic’ phase follower picking method with a small
method.
The procedure of signal elimination was executed prior to the manual selection of target
objects in order to eliminate the horizontal reflections, e.g., layers of the pavement structure. The f-k
filter was used in the model where for each individual profile, the parameter of the velocity range
of a fan-shaped form, with which the set objective was reached, was defined, since the procedure
was not satisfactory. A recognition analysis was done using a referential manual selection of the
target objects (see Figure 12), which are later used on migrated radargrams. The deconvolution was
additionally used on individual radargrams. It turned out to be useful in real urban site II in the
area of iron heating pipes, which cause visibly multiplying reflections. Pavement structure layers
Figure 12. Manual selection of target objects on processed non-migrated (red crosses and adjusted
Remote Sens. 2020,
estimations 12, x; made
were doi: FOR PEER REVIEW
using www.mdpi.com/journal/remotesensing
a ‘semi-automatic’ phase follower picking method with a small
hyperbolas) profile 14 of 400 MHz antennae. From left to right as follows: electronic cable duct (object
No. 14), meteor sewage (object No. 13), water supply (object No. 8), signal cable duct (object No. 15),
meteor sewage (object No. 13) and industrial water (object No. 9).
The procedure of signal elimination was executed prior to the manual selection of target objects
in order to eliminate the horizontal reflections, e.g., layers of the pavement structure. The f-k filter
was used in the model where for each individual profile, the parameter of the velocity range of a
fan-shaped form, with which the set objective was reached, was defined, since the procedure was not
satisfactory. A recognition analysis was done using a referential manual selection of the target objects
(see Figure 12), which are later used on migrated radargrams. The deconvolution was additionally
used on individual radargrams. It turned out to be useful in real urban www.mdpi.com/journal/remotesensing
Remote Sens. 2020, 12, x; doi: FOR PEER REVIEW
site II in the area of iron heating
pipes, which cause visibly multiplying reflections. Pavement structure layers estimations were made
using a ‘semi-automatic’ phase follower picking method with a small tolerance.
Figure 13a shows UUI on the discussed real urban site II according to the data of CCUI. Figure 13b
shows recorded UUI using the GPR-TPS. The cracks on the lined orthophoto recording, obtained using
the UVA (see appendix) due to the mending of the asphalt mixture when renovating or reconstructing
UUI (black lines along UUI), are very visible. The gas pipe (object No. 7) and the reconstructed water
pipe (object No. 8), which were recorded in the open trenches using geodetic methods, are in compliance
with the GPR-TPS model in terms of position and height. The rest of the UUI positionally stand
out quite significantly or do not even exist in the official recordings, which is attributed to historical
reasons, reconstructions and inconsistencies when recording. It can be seen that the UUI recorded
using the GPS-TPS model is positionally quite in compliance with visible traits when renovating or
reconstructing; it takes place through the middle of the excavation, which cannot be said for UUI
according to the data from CCUI.
methods, are in compliance with the GPR-TPS model in terms of position and height. The rest of the
UUI positionally stand out quite significantly or do not even exist in the official recordings, which is
attributed to historical reasons, reconstructions and inconsistencies when recording. It can be seen
that the UUI recorded using the GPS-TPS model is positionally quite in compliance with visible
traits
Remote when renovating
Sens. 2020, 12, 1228 or reconstructing; it takes place through the middle of the excavation, which
14 of 26
cannot be said for UUI according to the data from CCUI.
a) b)
Figure 13.
13. Recorded
Recordedunderground
underground utilities
utilities of urban
of real real urban siteTable
site II (see II (see Table
4): (a) 4): (a)toaccording
according to
stakeholders’
stakeholders'
data and CCUI; data
(b) and CCUI;to(b)
according dataaccording to data
collected based oncollected based
the proposed on the model.
GPR-TPS proposed GPR-TPS
model.
4. Results and Discussion
4. Results and Discussion
4.1. Real Urban Site I
4.1. Real
TheUrban Site I range, which is attributed to the characteristics of the subsurface and considerable
poor depth
moisture content and groundwater at depths of 1.2 to 1.5 m, was unified to all the processed GPR
The poor depth range, which is attributed to the characteristics of the subsurface and
profiles of the real urban site I. Consequently, the target objects were not detected on the expected
considerable moisture content and groundwater at depths of 1.2 to 1.5 m, was unified to all the
depths. Much better results were shown up to 1 m deep, where many well-expressed point reflections
processed GPR profiles of the real urban site I. Consequently, the target objects were not detected
can be detected.
on the expected depths. Much better results were shown up to 1 m deep, where many well-
The point reflection of the PVC pipe for the technological sewage in the far north of real urban
expressed point reflections can be detected.
site I, together with the existing object, can be seen extremely well on all the profiles (see Figure 14;
The point reflection of the PVC pipe for the technological sewage in the far north of real urban
meteor sewage (object No. 5); red cylindrical object). Similar can be said for the meteor sewage in the
site I, together with the existing object, can be seen extremely well on all the profiles (see Figure 14;
southern part (see Figure 14 (green cylindrical object)). According to the references of the investors
meteor sewage (object No. 5); red cylindrical object). Similar can be said for the meteor sewage in
and building constructors, other reflections could be attributed to the abandoned UUI, in particular
the southern part (see Figure 14 (green cylindrical object)). According to the references of the
the real urban site I. During reconstruction of the area, the building constructor did not remove the
investors and building constructors, other reflections could be attributed to the abandoned UUI, in
abandoned objects and pipes, hence their correct position and depth are unknown.
The sections of the studied volume of the ground, which provide a detailed insight into the
spatial relation of individual elements, are mostly used for the display of the results. This procedure is
especially welcome for the interactive interpretation in a 3D environment.
The assessment of the position accuracy and height described in Šarlah et al. [5] was executed
by taking the comparison of 12 contributing reference points, as shown in Figure 15. The assessed
positional accuracy of the GPR-TPS observations for the recorded UUI in the real urban site I is 0.146 m
(positional (radial) RMSEpos ). The positional (radial) standard deviation is σpos = 0.112 m, while the
centre of gravity of the vectors lays 0.098 m from the starting point in an easterly direction. The assessed
height accuracy is 0.134 m, while the σ∆H is 0.093 m and the central tendency of the height devitations
µ∆H = −0.128 m. The 3xRMSE threshold, which provides an observation of gross errors—i.e., an error
that was classified as an outlier - did not exceed any positional and height deviations. The difference
between the true incline and that obtained [5] using the GPR-TPS model was 0◦ 300 08” of the incline
angle, as shown in Figure 16.
particular the real urban site I. During reconstruction of the area, the building constructor did not
remove the abandoned objects and pipes, hence their correct position and depth are unknown.
The sections of the studied volume of the ground, which provide a detailed insight into the
spatial relation of individual elements, are mostly used for the display of the results. This procedure
Remote Sens. 2020, 12, 1228 15 of 26
is especially welcome for the interactive interpretation in a 3D environment.
Figure 14. Spatial relations among elements as presented using a series of parallel GPR-TPS profiles
of the real urban site I.
positional accuracy of the GPR-TPS observations for the recorded UUI in the real urban site I is
0.146 m (positional (radial) RMSEpos). The positional (radial) standard deviation is 𝜎 = 0.112 m,
while the centre of gravity of the vectors lays 0.098 m from the starting point in an easterly direction.
The assessed height accuracy is 0.134 m, while the 𝜎∆ is 0.093 m and the central tendency of the
height devitations 𝜇∆ = −0.128 m. The 3xRMSE threshold, which provides an observation of gross
errors - i.e. an error that was classified as an outlier - did not exceed any positional and height
FigureThe
deviations.
Figure 14. Spatial relations
14. Spatial relations
difference among
between
amongtheelements as
aspresented
true incline
elements and thatusing
presented a aseries
obtained
using [5]ofusing
series ofparallel
theGPR-TPS
parallelGPR-TPSprofiles
GPR-TPS model of
profiles
the real
was 0°of30' urban site I.
the08''
realofurban
the incline
site I. angle, as shown in Figure 16.
positional accuracy of the GPR-TPS observations for the recorded UUI in the real urban site I is
0.146 m (positional (radial) RMSEpos). The positional (radial) standard deviation is 𝜎 = 0.112 m,
while the centre of gravity of the vectors lays 0.098 m from the starting point in an easterly direction.
The assessed height accuracy is 0.134 m, while the 𝜎∆ is 0.093 m and the central tendency of the
height devitations 𝜇∆ = −0.128 m. The 3xRMSE threshold, which provides an observation of gross
errors - i.e. an error that was classified as an outlier - did not exceed any positional and height
deviations. The difference between the true incline and that obtained [5] using the GPR-TPS model
was 0° 30' 08'' of the incline angle, as shown in Figure 16.
a) b)
Figure15.
Figure Presentationofofthe
15.Presentation theposition
positionand
andheight
heightdifferences
differencesofofthe
theapex
apexpoints
pointsininthe
thereal
realurban
urban site I:
(a) Circles
sitedifferences(grey colour)
I: (a) Circles(brown mark
(grey colour) the
point);mark position differences
the position
(b) Line of
(brown)differences 0.10 m, 0.15 m,
of 0.10tendency
marks central 0.20 m
m, 0.15 m,of0.20 and 0.25
m and
height 0.25 0.146
m; m; 𝜇m
differences –
∆ =
RMSE
0.146 (blue
mpos–m.
−0.128 RMSE circle); 0.112
pos (blue m – σ0.112
circle); m – 𝜎 (brown
pos position circle);
position barycentre
(brown circle);ofbarycentre
vectors differences
of vectors(brown
point); (b) Line (brown) marks central tendency of height differences µ∆H = −0.128 m.
a) b)
Figure 15. Presentation of the position and height differences of the apex points in the real urban
site I: (a) Circles (grey colour) mark the position differences of 0.10 m, 0.15 m, 0.20 m and 0.25 m;
0.146 m – RMSEpos (blue circle); 0.112 m – 𝜎 position (brown circle); barycentre of vectors

Figure 16. Presentation of reference inclinations (black line; 2.3% inclination, r2 =2 0.95) and GPR-TPS
Figure 16. Presentation of reference inclinations (black line; 2.3% inclination, r = 0.95) and GPR-TPS
model (red line; 2.6% inclination, r2 = 0.51) of the meteor sewage (object No.5).
4.2. Real Urban Site II

Bad resolution as a consequence of the damping and dispersion of the EMW in the medium is
unified to some radargrams of the GPR profiles of the real urban site II. The cause is attributed to
the leakage of the water through the cracks in the asphalt surface layer of the pavement structure.
Remote Figure 16. 12,
Sens. 2020, Presentation
1228 of reference inclinations (black line; 2.3% inclination, r2 = 0.95) and GPR-TPS 16 of 26
4.2.
4.2. Real
Real Urban Site II
Urban Site II
Bad resolution
Bad resolution asas aa consequence
consequenceofofthe thedamping
damping and dispersion
and dispersion of the EMW
of the EMW in the medium
in the medium is is
unifiedto
unified tosome
some radargrams
radargrams of of the
the GPR
GPR profiles
profiles of the real urban site II. The cause
cause is attributed
attributed toto the
the leakage of the water through the cracks in the asphalt surface layer of the pavement
leakage of the water through the cracks in the asphalt surface layer of the pavement structure. When structure.
When laying,
laying, renovating
renovating and reconstructing
and reconstructing UUI inUUI
the in the discussed
discussed area, area, the asphalt
the asphalt layerslayers
werewere cut.
cut. Narrow
Narrow
and wideand wide
cracks crackson
appear appear on the discussed
the discussed areabe
area as can asseen
can be
in seen
Figure in 17.
Figure
The17. The insufficient
insufficient execution
and application of existing different asphalt mixtures probably affected the cracks alongalong
execution and application of existing different asphalt mixtures probably affected the cracks with the
with the reconstruction,
reconstruction, traffic and
traffic and climate. climate.
Filling up theFilling
cracksup the sense
in the cracksofin the of
input sense of input ofbitumen
the appropriate the
appropriate bitumen mixture during maintenance was not executed. Therefore, water enters the
mixture during maintenance was not executed. Therefore, water enters the pavement structure due to
pavement structure due to the porousness of the upper layers. GPR-TPS measurements were
the porousness of the upper layers. GPR-TPS measurements were executed following two days of rain,
executed following two days of rain, which is seen on the damping of EMW and the quality and
which is seen on the damping of EMW and the quality and resolution of individual radargrams, where
resolution of individual radargrams, where the water noticeably leaked through the cracks. Figure
the water noticeably leaked through the cracks. Figure 18 shows the cracks in the covering layer and
18 shows the cracks in the covering layer and the effect of water leakage into the subbase and base
the effect
course of water
layers and leakage
subgradeinto soilthe
on subbase and base
the radargram course
of the 25thlayers
profileand subgrade
of the 400 MHz soil on the radargram
antenna, where
of the 25th profile of the 400 MHz
the damping of EMW is the most obvious. antenna, where the damping of EMW is the most obvious.
a) b) c)
Figure 17.
Figure 17. Crevices
Crevices onon joints
jointsofofvarious
variousasphalt
asphaltmixtures
mixtures at at
thethe
time of reconstruction
time of reconstruction of pavement
of pavement
structure of
structure of real
realurban
urbansite
siteII:II:
(a)(a)
forfor
thethe
needs of the
needs ofheating system
the heating (no. 1) (no.
system and electronic cable ductcable
1) and electronic
(no. (no.
duct 2); (b)2);for
(b)the
for needs
the needs of the water
of the supply
water system
supply and and
system gas gas
pipeline (no.(no.
pipeline 3); (c)
3); visible
(c) visible
retention/accumulation
Sens. 2020, 12, x FOR PEERand
Remoteretention/accumulation REVIEW
and water
waterentry
entryafter
afterrainfall.
rainfall. 16 of 25
18. Visible
Figure 18.
Figure Visibleanomalies of the
anomalies of covering layer layer
the covering of the of
pavement structurestructure
the pavement caused bycaused
reconstruction
by
reconstruction (1—heating
(1—heating system, 2—electronic
system, 2—electronic
cable duct and 3—water supply
cable duct —water
and 3and gas pipeline)
supply andandgas
thepipeline)
consequences
of water
and penetration of
the consequences into the penetration
water pavement structure for the attenuation
into the pavement structure forofthe
EMW on profile
attenuation 25 of the
of EMW
on
400profile 25 of the 400 MHz antenna.
MHz antenna.
Consequently,the
Consequently, thefact
factofofapplicability
applicability ofof
thethe GPR-TPS
GPR-TPS model
model when
when determining
determining the position
the position of of
faults and damages on UUI and detecting the losses losses on the water
water supply
supply system
system can
can be
be concluded
concluded and
confirmed. When
and confirmed. discovering
When the losses,
discovering largelarge
the losses, relative water
relative permittivity
water and and
permittivity changes of the
changes of physical
the
physical characteristics
characteristics of the subsurface,
of the subsurface, which has which has been
been stated stated
by many by many
authors authors
[42–45], [42–45], was
was exploited. On the
exploited.
less On thepart
deteriorated less of
deteriorated
the pavement partstructure,
of the pavement
unburdenedstructure,
with unburdened
daily traffic, with
which daily traffic, by
is presented
which
the is presented
emergency laneby the emergency
(western laneintersection),
part of the (western partwater
of theleakage
intersection),
in the water leakage
pavement in the was
structure
pavement structure was not detected. Clearly visible reflections under the identical cracks
not detected. Clearly visible reflections under the identical cracks due to the patching of the asphalt due to
the patching of the asphalt mixture for the needs of the UUI reconstruction are shown in Figure
19a–c. Anomalies of the asphalt surface layer of the pavement structure due to the reconstruction
(1—heating system (objects No. 10, 11 and 12) , 2—electronic cable duct (object No. 14) and 3—
water supply system and gas line (objects No. 7 and 8) are visible. When processing radargrams,
deconvolution on the area of the iron gas pipes was also used. The application of deconvolution on
faults and damages on UUI and detecting the losses on the water supply system can be concluded
and confirmed. When discovering the losses, large relative water permittivity and changes of the
physical characteristics of the subsurface, which has been stated by many authors [42–45], was
exploited. On the less deteriorated part of the pavement structure, unburdened with daily traffic,
which is presented by the emergency lane (western part of the intersection), water leakage 17
Remote Sens. 2020, 12, 1228
in ofthe
26
pavement structure was not detected. Clearly visible reflections under the identical cracks due to
the patching of the asphalt mixture for the needs of the UUI reconstruction are shown in Figure
mixtureAnomalies
19a–c. for the needs of the
of the UUI reconstruction
asphalt surface layer ofare theshown in Figure
pavement 19a–c.due
structure Anomalies of the asphalt
to the reconstruction
surface layer of the pavement structure due to the reconstruction (1—heating
(1—heating system (objects No. 10, 11 and 12) , 2—electronic cable duct (object No. 14) and system (objects No.3— 10,
11 andsupply
water 12), 2—electronic
system andcable gas duct (object No.
line (objects No.14) and 8)
7 and 3—water supply
are visible. system
When and gasradargrams,
processing line (objects
No. 7 and 8) are
deconvolution on visible.
the areaWhen
of the processing radargrams,
iron gas pipes deconvolution
was also used. on theof
The application area of the iron gas
deconvolution on
pipes was also used. The application of deconvolution on the radargram of
the radargram of the 19th profile of the 400 MHz antenna, on which there are iron gas pipes that the 19th profile of the
400 MHz
cause antenna,
visible on which
multiplyaing there are iron
reflections, gas pipes
is shown. The that cause
interval of visible
startingmultiplyaing
and ending reflections,
double time, is
shown.determines
which The interval theofautocorrelation,
starting and ending was double
limited time,
from which
5 to 11determines thetraces
ns, filtering autocorrelation, was
on 10 ns, while
limited
the partfrom 5 toelimination
of the 11 ns, filtering traces
of the on 10noise
white ns, while
was theset part of the50.
at grade elimination of the white
Such a processed noise was
profile
set at grade 50.using
strengthened Such atheprocessed
function profile was strengthened
of manual signal gain using
for the function
better of manualFollowing
visualisation. signal gain thefor
better visualisation. Following the deconvolution, the reflection of the intake
deconvolution, the reflection of the intake and outtake of parallel heating pipes at an identical and outtake of parallel
heating
depth canpipes at an identical
be detected depth19).
(see Figure can be detected (see Figure 19).
a) b) c)
Figure 19. Visible
Figure 19. Visibleabnormalities
abnormalities of the
of the covering
covering layers
layers of GPR-TPS
of GPR-TPS data (1—heating
data (1—heating system, 2—
system, 2—electronic
electronic
cable ductcable
and 3—water 3—water
duct and and and gas(a)
gas supply): supply): (a) using deconvolution
using deconvolution on the processed
on the processed non-
non-migrated
migrated (interval
(interval 5–11 5–11
ns, 10 ns,ns, 10 ns, 50—
50—white white
noise) noise) radargram
radargram of19;
of profile profile 19; (b)deconvolution
(b) using using deconvolution
on the
on the processed
processed non-migrated
non-migrated radargramradargram of profile
of profile 14; (c) using14; (c) using deconvolution
deconvolution on the
on the processed processed
non-migrated
radargram of profile
non-migrated 9. of profile 9.
radargram
In CCUI, the steel pipes (objects No. 10 and 11) in the culvert (object No. 12) dimension 170 cm ×
80 cm (see Table 4) were recorded at the real urban site II. The sketch and dimensions of the closest
transverse profile (elaborate, no. 303/78-140 GZ) of the concrete plate and culvert with the built-in pipes,
measured using geodetic methods and a radargram of the 20th GPR profile, are shown in Figure 20.
From the sketch, the 14 cm thick concrete cover plate, under which there are just 4 cm on this part of
the route from the bottom of the heating pipes, can be presupposed. While presupposing the relative
permittivity of the concrete (4.5) and air (1) the referential double time from the top of the cover to
the apex point of the isolated pipe would be 2.24 ns. From Figure 21, the culvert dimension (171 cm)
and double time of the EMW propagation from the top of the cover of the culvert to the apex point of
isolated heating pipes (2.28 ns), based on the suggested GPR-TPS model by which the compliance of
the CCUI data and GPR-TPS observations can be confirmed. The radargram of the 20th profile was
only partly processed (signal delay elimination, manual determination of zero-time, spatial filter of the
background elimination, frequency band-pass filter, Kirchoff migration and repeated manual signal
gain) due to the explicit display of the reflection of the concrete cover of the heating pipes.
thethe culvert
culvert totothethe
apexapex point
point ofof isolated
isolated heating
heating pipes
pipes (2.28
(2.28 ns),
ns), based
based onon the
the suggested
suggested GPR-TPS
GPR-TPS
model
model byby which
which thethe compliance
compliance ofof
thethe CCUI
CCUI data
data and
and GPR-TPS
GPR-TPS observations
observations can
can bebe confirmed.
confirmed. TheThe
radargram of the 20th profile was only partly processed (signal
radargram of the 20th profile was only partly processed (signal delay elimination, manual delay elimination, manual
determination
determination ofof zero-time,
zero-time, spatial
spatial filter
filter ofof
thethe background
background elimination,
elimination, frequency
frequency band-pass
band-pass filter,
filter,
Kirchoff
Kirchoff migration
migration andand repeated
repeated manual
manual signal
signal gain)
gain) due
due to to the
the explicit
explicit display
display of of the
the reflection
reflection of
Remote Sens. 2020, 12, 1228 18of
of 26
the concrete cover of the heating
the concrete cover of the heating pipes. pipes.
a) a) b)b)
Figure
Figure 20.20.
Figure 20. CCUI
CCUI
CCUI data
data onon
data on heating
heating
heating system
system
system (objects
(objects
(objects No.
No. 10,10,
No. 1111
10, and
11 and
and12)12)
from
12) from
from 1988:
1988: a) a)
1988: aasketch
a sketch
(a) of of
sketch of the
the the
cross-sectional
cross-sectional
cross-sectional area
area
area of concrete
of concrete
of concrete slab
slab
slab and
and culvert
culvert
and with
with
culvert pipes
pipes
with pipes installed;
installed; b)
b) (b)
installed; the
thethe pipe
pipe type
type
pipe (JK)
(JK)(JK)
type and
andand radius
radius
radius
(ND/OD
(ND/OD
(ND/OD 250)
250) and
250) and
thethe
and dimension
dimension
the dimension of of
of the
the culvert
culvert
the (170
(170
culvert cmcm
(170 ××80
× 80
cm 80cm).
cm).cm).
a) a) b)b)
Figure21.21.
Figure
Figure GPR-TPS
GPR-TPS
GPR-TPS datadata
data about
about
about heating
heating
heating system
system (objects
(objects
(objects No.
No.No.
10, 10,
10,11 11
and
11 and
12):
and 12):
(a)
12): the(a)
(a) the
dimension
the dimension of of
of culvert
dimension
culvert
(171 cm)
culvert (171
(171 and cm)
cm) andand
double double
time
double oftimetimeof of
reflection reflection
(11.91
reflection (11.91
ns)(11.91
from ns)ns)
the from
apex
from ofthethe apex
concrete
apex of
cover
of concrete
of thecover
concrete cover
heating of
thethe
ofsystem
heating
culvert
heating system
on the
system culvert
basis
culvert onon
of thethe
the basis
of of
partially
basis the partially
processed
the partially processed
migrated migrated
radargram
processed migrated radargram
ofradargram
profile 20;of(b)of profile
20;20;
(b)(b)
double-time
profile of
double-time
double-time of of
reflection (14.18 reflection
ns) from
reflection (14.18
the
(14.18 ns)ns)
apex offrom
from thethe
insulated apex
apex of of
heating insulated heating
pipes using
insulated heating pipes
decompression
pipes using
using ofdecompression
the processedofand
decompression of
thethe processed
migrated
processed and
radargrams
and migrated
migrated radargrams
of profile 19.
radargrams of of profile
profile 19.19.
The degree of trust of the CCUI data in the heating system pipes (objects No. 10, 11 and 12) and
the Sens.
Remote electronic
Remote Sens. 2020,
2020, communication
12,12, x; doi:
x; doi: FORFOR duct
PEER
PEER (object No.
REVIEW
REVIEW 14) is low. It could bewww.mdpi.com/journal/remotesensing
argued that the positions obtained
www.mdpi.com/journal/remotesensing
according to the GPR-TPS model, are more correct. Patching of the asphalt mixture can be used as one
of the criteria for validation of the results of the heating and electronic communication network position.
The official position of the objective in CCUI does not match or positionally does not meet the position
of the width of the patching of the asphalt surface during reconstruction. When displaying the position
of officially valid CCUI data and that obtained using the GPR-TPS model, the positional correspondence
of the patched asphalt mixture and the GPR-TPS data, as shown in Figure 22, is concluded.
used as one
obtained of the to
according criteria for validation
the GPR-TPS model, of arethe results
more of the
correct. heating
Patching andasphalt
of the electronic communication
mixture can be
used as one
network of the criteria
position. for validation
The official position ofof the
theresults of the
objective heating
in CCUI andnot
does electronic
match communication
or positionally does
network
not meetposition. The official
the position of theposition
width of of the
theobjective
patchinginof
CCUI
the does not surface
asphalt match orduring
positionally does
reconstruction.
not meet the position of the width of the patching of the asphalt surface during
When displaying the position of officially valid CCUI data and that obtained using the GPR-TPS reconstruction.
When
model, displaying
the2020,
Remote Sens. 12, the
1228position
positional of officiallyofvalid
correspondence CCUI data
the patched and that
asphalt obtained
mixture andusing the GPR-TPS
the GPR-TPS data,
19 ofas
26
model,
shown the positional
in Figure 22, iscorrespondence
concluded. of the patched asphalt mixture and the GPR-TPS data, as
shown in Figure 22, is concluded.
Figure
Figure22.
Figure 22.Presentation
22. Presentation
Presentation ofofmatching
of matching
matching of of
repaired
of repaired
repairedasphalt in utility
asphalt
asphalt in infrastructure
in utility
utility renovation,
infrastructure
infrastructure GPR GPR
renovation,
renovation, GPR
measurements
measurementsand andofficial data
official onon
data thethe
CCUI
CCUI(see(see
Table 4).
Table 4).
Table 4).
The
Theposition
positionaccuracy
accuracy
accuracy andand
andheight
height
height assessment
assessment
assessment as described in Šarlah
as described
as described et al.et
in Šarlah
in Šarlah et[5]
al.were
al. executed
[5] were
[5] by by
executed
executed
comparing
comparing 4646 corresponding
46 corresponding reference
correspondingreference points
referencepoints of the
pointsofofthedrinking
drinking
the water
water
drinking supply
watersupply pipe
supply pipe (object
pipe No.
(object 7)
No.No.
(object and
7) and the
7) and
the
gasgas
the gas pipe
pipe
pipe (object
(object No.
(objectNo.
8),8),
No. 8),as
as shown
shown
as ininFigure
shown Figure
in 23.23.
23.
Figure
0,3
0,3
0,2
0,2
0,1
0,1
0.0
0.0
-0,1
-0,2 -0,1
-0,3 -0,2
-0,3 4 8 12 16 20 24 28 32 36 40 44
4 8 12 16 n 20 24 28 32 36 40 44
n
a) b)
a) b)
Figure 23. Position and height differences of apex points in a selected urban environment: (a) circles
Figure 23. Position
(grey colour) markand heightdifferences
position differences0.10
of apex points
m, 0.15 m,in a selected
0.20 urban
m and 0.25 m;environment: (a) circles
0.114 m - position RMSEpos
(grey colour) mark position differences 0.10 m, 0.15 m,points
0.20 minand 0.25 m; 0.114 menvironment:
- position RMSE
(blue circle); 0.111 m - position σpos (orange circle); the barycenter of vectors (orange point);
Figure 23. Position and height differences of apex a selected urban (a)pos(b)
circles
the
(blue circle); 0.111 m - position 𝜎 (orange circle); the barycenter of vectors (orange point); (b) the
(grey
orangecolour) markaposition
line marks differences
central tendency of 0.10 0.15 m, 0.20µm
m, differences
height = 0.003
∆Hand 0.25 m.
m; 0.114 m - position RMSEpos
orange line marks
(blue circle); 0.111a central tendency
m - position 𝜎 of(orange circle); the𝜇barycenter
height differences ∆ = 0.003 m.of vectors (orange point); (b) the
orange line marks a central tendency of height differences 𝜇
∆ = 0.003 m.
The assessed positional accuracy of the GPR-TPS observations when recording UUI at real
urban site II is 0.114 m (positional (radial) RMSEpos). The positional (radial) standard deviation σpos =
The assessed positional accuracy of the GPR-TPS observations when recording UUI at real
0.111 m, the centre of gravity of the deviations´ vectors, lies 0.038 m from the starting point in a
urban site II is 0.114 m (positional (radial) RMSEpos). The positional (radial) standard deviation σpos =
0.111 m, the centre of gravity of the deviations´ vectors, lies 0.038 m from the starting point in a
Remote Sens. 2020, 12, 1228 20 of 26
The assessed positional accuracy of the GPR-TPS observations when recording UUI at real urban
site II is 0.114 m (positional (radial) RMSEpos ). The positional (radial) standard deviation σpos = 0.111 m,
the centre of gravity of the deviations´ vectors, lies 0.038 m from the starting point in a southeasterly
direction. The assessed height accuracy is 0.100 m, while σ∆H equals 0.103 m and the central tendency
of the altitude deviations µ∆H = 0.003 m (see Figure 24). The central tendency expresses the height
movement of the target object according to the referential heights. A very good assessment of EMW
propagation spreading velocity is shown in combination with the method of hyperbola fitting and the
southeasterly direction. The assessed height accuracy is 0.100 m, while σ∆H equals 0.103 m and the
method of velocity determination based on individual layers of the pavement structure. None of the
central tendency of the altitude deviations μ∆H = 0.003 m (see Figure 24). The central tendency
positional and height deviations surpassed the threshold of the 3 × RMSE.
expresses the height movement of the target object according to the referential heights. A very good
Figure 24. 𝑅𝑀𝑆𝐸

RMSE𝑝𝑜𝑠 and𝑅𝑀𝑆𝐸
posand RMSE in in
∆𝐻∆H the
the3D3Dspace
spaceofofreal
realurban
urbansite
siteII.
II.The
The true
true position
position of
of water
water
supply
supply pipe
pipe (blue)
(blue) No. 77 and
and gas
gas pipe
pipe (yellow)
(yellow) No.
No. 88(see
(seeTable
Table 4)
4) and
and the
the corresponding
corresponding pipes
pipes (red)
(red)
determined on the basis of the GPR-TPS model.
Prior to the 3D-visualisation in three rectangular plains N, E and H for all the processed and
migrated radargrams, the envelopes were calculated using the Hilbert transformation, in which the
spatial relations between the chosen UUI can be seen more clearly.
An assessment of the incline accuracy of the UUI was carried out by comparing the true incline
and the incline obtained using the GPR-TPS model. By using linear regression, the best-fitting straight
line was found [5]. Two best-fitting straight lines per pipe are proposed due to the inclination which
varies as a result of joining of the two pipe sections. The difference between the true incline and that
obtained using the GPR-TPS model was 1◦ 240 03” of the incline angle, as shown in Figure 25.
a)
b)
Figure 24. 𝑅𝑀𝑆𝐸𝑝𝑜𝑠 and 𝑅𝑀𝑆𝐸∆𝐻 in the 3D space of real urban site II. The true position of water
supply pipe (blue) No. 7 and gas pipe (yellow) No. 8 (see Table 4) and the corresponding pipes (red)
Remote Sens. 2020, 12,
determined on 1228
the basis of the GPR-TPS model. 21 of 26
a)
b)
Figure
Figure25.25.Inclinations
Inclinationsofofreference
reference(black
(blackline)
line)andandproposed
proposedGPR GPRmodel
model(red(redline):
line):(a)
(a)black
blackline
line
section
sectionI:I:2.8% inclination,r2r=2 =
2.8%inclination, 0.99; black
0.99; line
black linesection II: II:
section 0.2% inclination,
0.2% inclination, = 0.52;
r2 =r20.52; redred
lineline
section I:
section
3.6% inclination,
I: 3.6% inclination, = 0.86;
r2 =r20.86; redred
line section
line II: II:
section 0.7%
0.7% inclination,r2 r=2 =
inclination, 0.26
0.26water
watersupply
supplypipes
pipesNo.
No.77
(RMSE∆H = 0.072 m, µ∆H = 0.063 m, σ∆H = 0.032 m, RMSEpos = 0.048 m, σpos = 0.05 m); (b) black line
Remote section
Sens. 2020, 12, x; inclination,
I: 1.3% doi: FOR PEER r2 REVIEW
= 0.76; black line section II: 0.2% inclination, www.mdpi.com/journal/remotesensing
r2 = 0.10; red line section I:
3.5% inclination, r = 0.61; red line section II: 0.6% inclination, r = 0.16 gas pipe No. 8 (RMSE∆H =
2 2
0.128 m, µ∆H = −0.058 m, σ∆H = 0.113 m, RMSEpos = 0.116 m, σpos = 0.152 m).
5. Conclusions
Through this work, the authors have presented a suitable solution of the kinematic
ground-penetrating radar (GPR) and self-tracking (robotic) terrestrial positioning system (GPR-TPS)
model for the acquisition, processing and visualisation of underground utility infrastructure (UUI)
with high 3D-georeferenced accuracy in a real urban environment. The kinematic GPR-TPS model
used for this research was created for high-accuracy underground utilities’ 3D mapping in real urban
field/testing pools [5]. Only minor adaptations were necessary for the transfer to fields of GPR
investigation in real urban environments. The GPR-TPS model meets the recommended positional
and height accuracy of second quality levels (QL-B) and gets close to first quality levels (QL-A) of
the recognised UUI. By using the GPR-TTP model, the positional and height accuracy was less than
15 cm in the real urban sites I and II (RMSEpos = 14.6 cm; RMSE∆H = 13.4 cm and RMSEpos = 11.4 cm;
RMSE∆H = 10.1 cm). The model proved to be a useful tool in determining the inclination of UUI, which
is restricted by the height accuracy of the method (RMSE∆inc = 0.5◦ and RMSE∆inc = 1.5◦ ). For better
presentation of the results of UUI obtained using the GPR-TPS model, an additional recording using
small UAV was executed. The UAV recording turned out to be extremely useful when confirming the
Remote Sens. 2020, 12, 1228 22 of 26
actual position of the UUI, with the help of detected patches of the asphalt surfaces on the orthophoto.
For this reason, the authors propose that the best way to map UUI is to cross both non-destructive
techniques: the UAV photogrammetry to cover the surface-specific information, and the GPR-TPS
model to cover subsurface information about UUI.
One of the obvious advantages of the kinematic GPR-TPS model for mapping the UUI is that it
is not necessary to establish an orthogonal grid of the researched area, on which subsequently the
lengths of the measured GPR profiles are usually linearised. The second obvious advantage of the
suggested model is that, in compliance with the density of the contained positional and height points
in real-time mode, the path of the movement of the GPR antenna presents a more credible position
of the measured profile. The third advantage is in capturing accurate and continuous heights along
the entire GPR trajectory which, with all the small changes in the shape of the surface, are key for the
execution and research goals for exact topographic correction.
Of course, the GPR method has some limitations in recognitions and recordings of the UUI.
Questions and gaps arising with the kinematic GPR-TPS model will have to be improved in terms of
implementation of approaches to automatically identify and classify hyperbolic signatures. The success
of the GPR-TPS model always depends on many aggravating circumstances, which appear in the
real urban environment [1–3]. The example of the selected real urban site I, where the damping
and dispersion of the EMW are caused due to the high moisture content or the large coarse ground
tampon in the medium, proves that reflections—and with them the recognisability of the UUI—are
fundamentally diminished. Similar conclusions also apply for the real urban site II, on which, due to
the asphalt patches when reconstructing the UUI and consequently inappropriate maintenance, water
has entered the road body. In this part, the extremely aggravating circumstances showed as a presence
of moisture in the subsurface, and in some cases also to the extreme that the GPR method is unreliable
or even completely useless.
It is presupposed that the results of the presented GPR-TPS model at the level of the development,
as has been shown in recent years, have great potential in many areas. The social consequences of the
suggested model are seen in the areas of spatial planning, obtaining building permission, protecting
and insuring underground property, preventing damage, etc. Damage claims due to direct and indirect
damage caused to UUI are also questionable. The first consequence, which can very quickly occur in
the Republic of Slovenia, is that the investor or the stakeholder of the underground UUI in the phase
of planning the intervention in the space or obtaining the adequate documentation (e.g., building
permission), when there is insufficient knowledge or bad positional accuracy of the existing UUI in
the area of the intervention, would have to record the sub-surface of the existing UUI, which would
serve as a foundation for projections. This brings common conflicts, costs and material damage in
spaces which occur due to not knowing the true position of the existing UUI. Nowadays, this is almost
common practice in spatial areas or those of the legal regime of cultural heritage protection, where the
GPR method is widely used in terms of previous archaeological studies of cultural heritage, which
is ultimately cheaper than interventions in the spaces themselves. It is an encouraging fact that the
GPR-TPS method is financially cheaper than digging interventions and repeating reconstructions of
urban areas, especially when considering the fact that the position of the route of the UUI stands as an
unknown when the intervention occurs.
Finally, through this work the authors have demonstrated that by using the suggested GPR-TPS
model objectively it is possible to determine the quality (especially the parameter of accuracy) of the
information obtained about the UUI in real urban area.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-4292/12/8/1228/s1,

Video S1: Introduction video.
Author Contributions: N.Š., T.P. and B.M. conceived, designed and performed the research; N.Š. and B.M.
analysed the data and discussed the results; T.A. contributed adjustment; N.Š. wrote the manuscript; T.P., T.A. and
B.M. reviewed the manuscript and approved the final version. All authors have read and agreed to the published
version of the manuscript.
Remote Sens. 2020, 12, 1228 23 of 26
Funding: This research received no external funding.

Acknowledgments: The authors would like to thank the anonymous reviewers for their valuable critiques and
suggestions which improved the article.
Conflicts of Interest: The authors declare no conflict of interest.
Appendix A. Unmanned Aerial Vehicle Acquisition and Processing

A complete photogrammetry workflow was used to generate an accurate orthomosaic with a
spatial resolution of 1.17 cm on a surface of 2.56 ha with UAV. Indirect geo-referencing was applied
using the measuring reference point that was deployed on the ground in the real urban site II before
the flight. Nine ground control points, which were used for orientation of the recording and indirect
geo-referencing of the photogrammetric block, were equally spread, stabilised and signalised across
the area. Four points of the basic poligonometric net of the I. order, which are based on a rough
positional and height assessment of the accuracy of the DTM, were used as the check points. Classic
terrestrial measurement and the GNSS method were used when determining the horizontal position
and the height of the ground control points and check points. The recording using UAV was executed
following the execution of the geodetic measurements and determination of the area of the flight.
The UAV recording was taken with an 80% longitudinal overlap, a 40% lateral overlap, a flying altitude
of 90 m, a 36-megapixel resolution (7340 × 4912), a 35 mm fixed focal length, and an average pixel
(4.88 µm × 4.88 µm).
Joining the images into a photogrammetric block takes place through the tie points, which are
automatically determined using dense image matching. In this phase of the processing, where the
data only connect relatively with each other, the elements of the interior orientation of the camera
(coordinates of the main point of the recording, the focal length, the size of the pixel and the parameters
of radial and/or tangential optical distortion) are also calculated. The parameters of interior orientation
for the cameras were determined using laboratory calibration [46]. The relatively oriented block of
recording which is based on the coordinates of ground control points (see Figure A1), was transformed
into the referential coordinate system. Meanwhile, the parameters of the exterior orientation of each
recording in the block was also calculated. The photogrammetric point cloud was calculated in
the adjustment
Remote using
Sens. 2020, 12, x FORoriented recordings and automatically measured image coordinates of the
PEER REVIEW 22 oftie
25
points [47]. The cloud of points is later classified on the ground and non-ground points. From the
ground points,
ground and non-ground
the DTM whichpoints.serves
Fromtothe ground points,
orthorectify the DTM is
the recordings which servesusing
calculated to orthorectify the
interpolation.
recordings
The is calculated
final orthophoto using
of the realinterpolation. Thetaken
urban site II was final orthophoto of the real
using a mosaicking urban
of the site II waswhich
orthophotos, taken
usingtaken
were a mosaicking
from the of the orthophotos,
singular recordings.which were taken from the singular recordings.
a) b)
Figure A1.
Figure A1. (a)
(a)Position
Positionofofthe
the ground
ground control
control (yellow)
(yellow) andand check
check points
points (green)
(green) ofrecording
of the the recording
area;
area; (b) overlapping 30 images’ frequency and camera
(b) overlapping 30 images’ frequency and camera position. position.
Checking
Checking the
theaccuracy
accuracyofofDTM
DTMdata in in
data 3D3D
space is carried
space out out
is carried by comparing the DTM
by comparing data with
the DTM data
reference valuesvalues
with reference on checkpoints. The reference
on checkpoints. values of
The reference the checkpoints
values were determined
of the checkpoints in the field
were determined in
the field using geodetic methods (GNSS) with very high accuracy. The accuracy of DTM is
expressed using vertical accuracy measures and horizontal (planimetric) accuracy measures [48,49].
The distribution and number of checkpoints, in this case, was not satisfactory for an actual quality
assessment, however, the accuracy of the orthophoto was at least approximately defined. Since the
normal distribution of errors was not proved, robust accuracy measures were used, namely: 50%
Remote Sens. 2020, 12, 1228 24 of 26
using geodetic methods (GNSS) with very high accuracy. The accuracy of DTM is expressed using
vertical accuracy measures and horizontal (planimetric) accuracy measures [48,49]. The distribution
and number of checkpoints, in this case, was not satisfactory for an actual quality assessment, however,
the accuracy of the orthophoto was at least approximately defined. Since the normal distribution of
errors was not proved, robust accuracy measures were used, namely: 50% quantile Q|∆i|(0.50) and
95% quantile Q|∆i|(0.95) of the absolute values of the errors and normalised median absolute deviation
(NMAD). Based on the robust accuracy measures Q|∆H|(0.95) = 5.0 cm it was assessed that the height
accuracy of DTM is: NMAD = 2.2 cm, Q|∆i|(0.50) = −1.4 cm; the positional accuracy of the orthophoto
on the coordinate axis y Q|y|(0.95) = 1.7 cm (NMAD = 1.2 cm, Q|∆i|(0.50) = 0.003 cm) and coordinate
axis x Q|x|(0.95) = 4.7 cm (NMAD = 2.7 cm, Q|∆i|(0.50) = −0.007 cm). The great difference between
the coordinate axes is attributed to the extremely low number of control points. The robust accuracy
measure of the orthophoto according to the horizontal position of Q|xy|(0.95) was 4.9 cm.
References
1. Annan, A.P. Ground Penetrating Radar: Principles, Procedures & Applications; Sensors & Software Inc.:
Mississauga, ON, Canada, 2003.
2. Jol, H.M. Ground Penetrating Radar: Theory and Applications; Elsevier Science: Amsterdam, The Netherlands;
Oxford, UK, 2009.
3. Persico, R. Introduction to Ground Penetrating Radar: Inverse Scattering and Data Processing; Wiley-IEEE Press:
Hoboken, NJ, USA, 2014.
4. Wai-Lok Lai, W.; Dérobert, X.; Annan, P. A review of Ground Penetrating Radar application in civil
engineering: A 30-year journey from Locating and Testing to Imaging and Diagnosis. NDT E Int. 2018, 96,
58–78. [CrossRef]
5. Šarlah, N.; Podobnikar, T.; Mongus, D.; Ambrožič, T.; Mušič, B. Kinematic GPR-TPS Model for Infrastructure
Asset Identification with High 3D Georeference Accuracy Developed in a Real Urban Test Field. Remote Sens.
2019, 11, 1457. [CrossRef]
6. Ayala-Cabrera, D.; Herrera, M.; Izquierdo, J.; Ocaña-Levario, S.J.; Pérez-García, R. GPR-Based Water Leak
Models in Water Distribution Systems. Sensors 2013, 13, 15912–15936. [CrossRef]
7. Metje, N.; Atkins, P.R.; Brennan, M.J.; Chapman, D.N.; Lim, H.M.; Machell, J.; Muggleton, J.M.; Pennock, S.;
Ratcliffe, J.; Redfern, M.; et al. Mapping the underworld—State-of-the-art review. Tunn. Undergr. Sp. Tech.
2007, 5–6, 568–586. [CrossRef]
8. Mohar Bastar, K.; Šarlah, N. The influence of the cost reduction directive on the broadband roll. Teh. Vjesn.
2019, 96, 58–78.
9. Šarlah, N. Exchange Formats for Interoperability Support of the Consolidated Cadastre of Public Infrastructure
Information System. Master’s Thesis, University of Ljubljana, Faculty of Civil and Geodetic Engineering,
Ljubljana, Slovenia, 2008. Available online: https://repozitorij.uni-lj.si/Dokument.php?id=84240&lang=slv
(accessed on 27 July 2019).
10. ASCE CI/ASCE38-02 American Society of Civil Engineers; Standard Guidelines for the Collection and
Depiction of Existing Subsurface Utility Data. ASCE 2002, 38, 1–20.
11. AS 5488—2013. Australian Standard: Classification of Subsurface Utility Information. 2013. Available online:
https://nulca.com.au/db_uploads/5488-2013_V2.pdf (accessed on 21 April 2018).
12. British Standars Institutions. British Standard: Specification for Underground Utility Detection, Verification
and Location Institute of Civil Engineering; ICE PAS 128: 2014; BSI: London, UK, 2014; Available online:
https://shop.bsigroup.com/ProductDetail/?pid=000000000030267400 (accessed on 1 April 2020).
13. Surveying and Mapping Authority of the Republic of Slovenia. Exchange Format and the Code Lists of
the Consolidated Cadastre of Utility Infrastructure. 2018. Available online: http://www.e-prostor.gov.si/
fileadmin/GJI/elaborati/Format_sifrant_6.02.pdf (accessed on 19 May 2019).
14. Šarlah, N.; Kumer, J.; Kuzmič, M.; Jud, S.; Mesner, A.; Mlinar, J.; Klemenčič, G.; Šelek, L. Guideline for
Underground Utility Mapping in Slovenia 2010. Available online: https://www.researchgate.net/publication/
263949930_GUIDELINE_FOR_UNDERGROUND_UTILITY_MAPPING_IN_SLOVENIA (accessed on 19
May 2019).
Remote Sens. 2020, 12, 1228 25 of 26
15. Grandjean, G.; Gourry, J.C.; Bitri, A. Evaluation of GPR techniques for civil-engineering applications: A
study on a test site. J. Appl. Geophys. 2000, 45, 141–156. [CrossRef]
16. Mušič, B.; Podobnikar, T.; Šarlah, N. Recording of Buried Utility Pipes with a Non-Invasive
Ground-Penetrating Method. Geographic Information System in Slovenia 2011–2012. 2012. Available
online: https://www.academia.edu/2519279/RECORDING_OF_BURIED_UTILITY_PIPES_WITH_NONINVASIVE_
GROUND_PENETRATING_METHOD (accessed on 30 May 2019).
17. Aaltonen, J.; Nissen, J. Geological mapping using GPR and differential GPS positioning: A case. In Proceedings
of the Ninth International Conference on Ground Penetrating Radar, Santa Barbara, CA, USA, 29 April–2
May 2002.
18. Jaw, S.W.; Hashim, M. Locational accuracy of underground utility mapping using ground-penetrating radar.
Tunn. Undergr. Space Technol. 2013, 35, 20–29. [CrossRef]
19. Li, S.; Cai, H.; Kamat, V.R. Uncertainty-aware geospatial system for mapping and visualising underground
utilities. Autom. Constr. 2015, 53, 105–119. [CrossRef]
20. Boniger, U.; Tronicke, J. On the Potential of Kinematic GPR Surveying Using a Self-Tracking Total Station:
Evaluating System Crosstalk and Latency. IEEE Trans. Geosci. Remote Sens. 2010, 48, 3792–3798. [CrossRef]
21. Böniger, U.; Tronicke, J. Improving the interpretability of 3D GPR data using target-specific attributes:
Application to tomb detection. J. Archaeol. Sci. 2010, 37, 672–679. [CrossRef]
22. Allroggen, N.; Tronicke, J.; Delock, M.; Böniger, U. Topographic migration of 2D and 3D ground-penetrating
radar data considering variable velocities. Near Surf. Geophys. 2014, 13, 253–259. [CrossRef]
23. Bilal, M.; Khan, W.; Muggleton, J.; Rustighi, E.; Jenks, H.; Pennock, S.R.; Atkins, P.R.; Cohn, A. Inferring the
most probable maps of underground utilities using Bayesian mapping model. J. Appl. Geophys. 2018, 150,
52–66. [CrossRef]
24. Dou, Q.; Wei, L.; Magee, D.R.; Atkins, P.R.; Chapman, D.N.; Curioni, G.; Goddard, K.F.; Hayati, F.; Jenks, H.;
Metje, N.; et al. 3D Buried Utility Location Using A Marching-Cross-Section Algorithm for Multi-Sensor
Data Fusion. Sensors 2016, 16, 1827. [CrossRef] [PubMed]
25. Chen, H.; Cohn, A.G. Buried utility pipeline mapping based on multiple spatial data sources: A Bayesian data fusion
approach. In Proceedings of the Twenty-Second International Joint Conference on Artificia Intelligence, Barcelona,
Spain, 16–22 July 2011; Available online: https://www.researchgate.net/publication/220815962_Buried_Utility_
Pipeline_Mapping_Based_on_Multiple_Spatial_Data_Sources_A_Bayesian_Data_Fusion_Approach (accessed on
20 March 2020).
26. Cheng, N.-F.; Tang, H.-W.; Chan, C.-T. Identification and positioning of underground utilities using
ground-penetrating radar (GPR). Sustain. Environ. Res. 2013, 23, 141–152.
27. Gabryś, M.; Kryszyn, K.; Ortyl, Ł. GPR surveying method as a tool for geodetic verification of GESUT
database of utilities in light of BSI PAS128. Rep. Geod. Geoinf. 2019, 107, 49–59. [CrossRef]
28. Ghozzi, R.; Lahouar, S.; Besbes, K.; Souani, C. Mapping of Sewer Lines Using GPR: A Case Study in Tunisia.
Data 2018, 3, 40. [CrossRef]
29. Ismail, N.A.; Saad, R.; Muztaza, N.M.; Ali, N. Predictive mapping of underground utilities using
ground-penetrating radar. Casp. J. Appl. Sci. Res. 2013, 2, 104–108.
30. Jeng, Y.; Chen, C.-S. Subsurface GPR imaging of a potential collapse area in urban environments. Eng. Geol.
2012, 147-148, 57–67. [CrossRef]
31. Metwaly, M. Application of GPR technique for subsurface utility mapping: A case study from an urban area
of Holy Mecca, Saudi Arabia. Measurement 2015, 60, 139–145. [CrossRef]
32. Porsani, J.L.; Ruy, Y.B.; Ramos, F.P.; Yamanouth, G.R.B. GPR applied to mapping utilities along the route
of Line 4 (yellow) subway tunnel construction in São Paulo City, Brazil. J. Appl. Geophys. 2012, 80, 25–31.
[CrossRef]
33. Sagnard, F.; Norgeot, C.; Derobert, X.; Baltazart, V.; Merliot, E.; Derkx, F.; Lebental, B. Utility detection and
positioning on the urban site Sense-City using Ground-Penetrating Radar systems. Measurement 2016, 88,
318–330. [CrossRef]
34. Van der Kruk, J. Three-Dimensional Imaging of Multi-Component GPR Data. Ph.D. Thesis, Delft University
of Technology, Delft, The Netherlands, 2001. Available online: http://e-collection.library.ethz.ch/eserv/eth:
24077/eth-24077-01.pdf (accessed on 13 March 2019).
35. Leckebusch, J. Ground Penetrating Radar: A Modern Three-dimensional Prospection Method. Archaeol.
Prospect. 2003, 10, 213–240. [CrossRef]
Remote Sens. 2020, 12, 1228 26 of 26
36. Luo, T.X.; Lai, W.W.; Chang, R.K.; Goodman, D. GPR imaging criteria. J. Appl. Geophys. 2019, 165, 37–48.
[CrossRef]
37. Kirschner, H.; Stempfhuber, W. The Kinematic Potential of Modern Tracking Total Stations—A State of the
Art Report on the Leica TPS1200+. In Proceedings of the 1st International Conference on Machine Control &
Guidance, Kinematic Measurement and Sensor Technology I (Local Systems), Zurich, Switzerland, 24–26
June 2008; pp. 51–60.
38. Leica TPS1200+. User Manual; Leica Geosystems: Heerbrugg, Switzerland, 2008; pp. 1–209. Available
online: http://webarchiv.ethz.ch/geometh-data/student/anleitungen/leicaman/tps1200/TPS1200_User_en.pdf
(accessed on 14 April 2018).
39. Siebert, S.; Teizer, J. Mobile 3D mapping for surveying earthwork projects using an Unmanned Aerial Vehicle
(UAV) system. Automat. Constr. 2014, 41, 1–14. [CrossRef]
40. Fraga-Lamas, P.; Ramos, L.; Mondéjar-Guerra, V.; Fernández-Caramés, T.M. A Review on IoT Deep Learning
UAV Systems for Autonomous Obstacle Detection and Collision Avoidance. Remote Sens. 2019, 11, 2144.
[CrossRef]
41. Bianchini Ciampoli, L.; Tosti, F.; Economou, N.; Benedetto, F. Signal Processing of GPR Data for Road Surveys.
Geosciences 2019, 9, 96. [CrossRef]
42. Hyun, S.-Y.; Jo, Y.-S.; Oh, H.-C.; Kim, S.-Y.; Kim, Y.-S. A laboratory scaled-down model of ground-penetrating
radar for leak detection of water pipes. Meas. Sci. Technol. 2007, 18, 2791–2799. [CrossRef]
43. Huisman, J.A.; Hubbard, S.S.; Redman, J.D.; Annan, A.P. Measuring Soil Water Content with Ground
Penetrating Radar: A Review. Vadose Zone J. 2003, 2, 476–491. [CrossRef]
44. Liu, G.; Jia, Y.; Liu, H.; Qiu, H.; Qiu, D.; Shan, H. A Case Study to Detect the Leakage of Underground
Pressureless Cement Sewage Water Pipe Using GPR, Electrical, and Chemical Data. Environ. Sci. Technol.
2002, 36, 1077–1085. [CrossRef]
45. Stampolidis, A.; Soupios, P.; Vallianatos, F.; Tsokas, G.N. Detection of leaks in buried plastic water distribution
pipes in urban places—A case study. In Proceedings of the 2nd International Workshop on Advanced
Ground Penetrating Radar, Delft, The Netherlands, 14–16 May 2003; pp. 120–124.
46. Kraus, K. Photogrammetrie, Band 1, Geometrischs Informationen aus Photographien und Laserscanneraufnahmen;
Walter de Gruyter: Berlin, Germany, 2009.
47. Rosnell, T.; Honkavaara, E. Point Cloud Generation from Aerial Image Data Acquired by a Quadrocopter
Type Micro Unmanned Aerial Vehicle and a Digital Still Camera. Sensors 2012, 12, 453–480. [CrossRef]
[PubMed]
48. Höhle, J.; Potuckova, M. EuroSDR No. 60: Assessment of the Quality of Digital Terrain Models. Off. Publ.
Eur. Spat. Data Res. 2009, 60, 1–85.
49. Höhle, J.; Höhle, M. Accuracy assessment of digital elevation models by means of robust statistical methods.
ISPRS J. Photogramm. 2009, 64, 398–406. [CrossRef]
© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).

Remotesensing 12 01228 v2 PDF

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Remotesensing 12 01228 v2 PDF

Caricato da

Copyright:

Formati disponibili

remote sensing

Keywords: kinematic GPR-TPS model; self-tracking terrestrial positioning system; underground

Remote Sens. 2020, 12, 1228; doi:10.3390/rs12081228 www.mdpi.com/journal/remotesensing

Table 2. Some examples of applications of ground-penetrating radar on underground utility

Remote Sens. 2020, 12, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing

Object Nominal Diam. Real Urban

3. Methods and Instrumentation

Remote Sens. 2020, 12, x FOR PEER REVIEW 7 of 25

Figures 5–8 and Table 5 show show the

Remote Sens. 2020, 12, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing

3.2. Real Urban Site II

Remote Sens. 2020, 12, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing

Remote Sens. 2020, 12, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing

4.2. Real Urban Site II

Remote Sens. 2020, 12, x; doi: FOR PEER REVIEW www.mdpi.com/journal/remotesensing

Figure 24. 𝑅𝑀𝑆𝐸

Supplementary Materials: The following are available online at http://www.mdpi.com/2072-4292/12/8/1228/s1,

Funding: This research received no external funding.

Appendix A. Unmanned Aerial Vehicle Acquisition and Processing

Potrebbero piacerti anche