DOI: https://doi.org/10.26682/sjuod.2020.23.1.

2
Journal of University of Duhok, Vol. 32, No.1 (Pure and Eng. Sciences), Pp 11-24, 3232

EVALUATION OF UAV-BASED DEM FOR VOLUME CALCULATION

AGEED IDREES M*, and FARSAT HEETO ABULRAHMAN**

*
Dept. of civil Engineering, College of Engineering, University of Duhok, Kurdistan Region-Iraq
**
Dept. of Surveying Engineering, College of Engineering, University of Duhok, Kurdistan Region-Iraq

(Received: December 10, 2019; Accepted for Publication: February 5, 2020)

ABSTRACT
In the latest decades, Unmanned Aerial Vehicles (UAVs) have witnessed rapid growth and it plays a vital
role in different fields of engineering and architecture. This technique can also be applied in land surveying
as a device in order to measure the 3D ground coordinates and Earth work. The main aim of this paper is
to evaluate the accuracy of volume that obtained by using Digital Elevation Model (DEM) derived from
UAVs images.
In this research, three different flights were performed with DJI phantom 4 pro (25m, 50m, and 100m)
with 80% forward and side overlaps at Duhok Dam. Several Ground control points (GCPs) were installed
and evenly distributed throughout the study area and their coordinates were determined using GPS-RTK
technique for geo-referencing. The data images captured with UAV were processed using Agisoft photoscan
Professional software. GPS survey was carried out using Leica viva GS10 base, and GS15 rover for the same
place. The volumes acquired by the UAV images including all three flight heights were compared to the
volume obtained with GPS survey techniques which considered as a base for comparison. The results
showed that the volume calculated with UAV images encountered to the base were compatible with each
other with (99. 86%,99. 76% and 99. 74%) for altitudes (25,50, and 100) respectively.

KEYWORD: UAV images, accuracy assessment of DEM, volume commutation, GCP, GPS-RTK.
https://doi.org/10.26682/sjuod.2020.23.1.2

1. INTRODUCTION A Digital Elevation Model (DEM) is an

important product that represents topography of

T he rapid development of Unmanned

Aerial Vehicles (UAVs) contributed in
various disciplines of engineering and
the ground surface or defines the Z values of
ground surfaces digitally with different
accuracies for variety of application fields (Polat
architecture. They are also used as a device to & Uysal, 2018). Also there are other two most
measure 3D ground coordinates and compute widely used terms in the literature: digital surface
earthworks which can be applied in land model (DSM) and digital terrain model (DTM).
surveying. With the help of photogrammetry The terms DEM, DTM, and DSM are generally
fundamental rules, UAV system is used to study used to denote to diverse types of continuous and
the Earth’s surface. Aerial surveys are becoming three-dimensional (3D) geospatial data. A DEM
more and more popular as a result of high spatial is represented as a three-dimensional (3D) raster
resolution of the images captured from UAV’s. image that shows the elevations of ground above
UAVs seem to be more efficient than ground the mean sea level by its pixel values. A DSM is
surveys with advantage of time and flexibility defined as a raster image that represents the
compared to the traditional surveying such as elevations of ground above the mean sea level by
total station in the small areas of study, and they its pixel values including all features such as hills,
have lower cost than conventional trees, buildings, etc., present on it, whereas a
photogrammetric flights using expensive metric DTM is a raster image that represents the
digital cameras mounted on the board or LiDAR elevations of ground above the mean sea level
sensors (Zietara, 2017) Images captured from (MSL)or above vertical datum by its pixel values
(UAVs) are used for generating Digital Surface as shown in figure (1) (Ravi, 2018) (Ajayi ,
Models (DSMs) and orthophoto (Saskia , Ruedi , Salubi, Angbas , & Odigure, 2017).
& Daniel, 2017).

11
ageed.mohammed@uod.ac
 Journal of University of Duhok, Vol. 32, No.1 (Pure and Eng. Sciences), Pp 11-24, 3232

Fig. (1):- Difference between DSM and DTM (BEGASHAW, 2018).

In recent years, UAVs have also been used for on the assessment of DEMs derived from both
volume calculation for different purposes such as UAV and Global Navigation Satellite System
excavation and filling of mines, huge construction (GNSS) for earthwork computations.
sites, the places where coal dumped and recycling Furthermore, (Tucci, Gebbia, Conti, Fiorini, &
area. In addition, activities of road construction Lubello, 2019) used UAV system for monitoring
are treated as one of the most economic and computing stockpile volumes of bulk
operations, since the cost of earthwork has materials etc., therefore the accuracy of DEM is
commonly the major portion among these an important subject to be studied in volume
activities (Akgul, Yurtseven, Gulci, & Akay, calculation. The main aim of this paper is to
2017). Undoubtedly, it is also possible to evaluate the accuracy of volume that obtained by
calculate volumes by using conventional geodetic using Digital Elevation Model (DEM) derived
methods including Theodolite, total station and from UAVs images at different flight heights.
GPS, but by taking the time saving, accessibility 2. RELATED WORK
and cost efficiency into account, UAV is
considered the ideal solution(Ulvi̇ , 2018). Numerous most recent researches and
investigations have written and published on the
The major concern that is related to the accuracy volume calculation derived from UAVs
accuracy of DEM should be taken into using variety procedures and software.
consideration and the main thing to take into (Wang, Al-Shabbani, Sturgill, Kirk, & Dadi,
account is whether the DSM produced from 2017), estimated earthwork volume of a stockpile
accuracy of aerial surveys implemented by UAV and trench through the use of conventional GPS
is sufficiently accurate to be used in volume survey and unmanned aerial system (UAS). The
computation. flight was performed with help of dual-controlled
Over recent years, many investigations have DJI Inspire 1 drone with 75% forward overlap
been done to evaluate the quality of products and 60% side overlap. The captured images were
generated from data collected by drones. For processed with Pix4Dmapper software, and
instance, (Ulvi̇ , 2018), analyzed the utility of comparison was made between volumes obtained
unmanned aerial vehicle in volume calculation, by both methods. The result of their study showed
while (Akgul et al., 2017) conducted a research that the error in volume for stockpile and trench

12
ageed.mohammed@uod.ac
in comparison with GPS survey were found about was made, and the results showed an agreement
±206.94 ft3 (7%) and ±3311.4563 ft3 (0.9%) between both methods within 3 755 m3 (0.7%) of
respectively. the total volume or 5 cm thickness over the entire
surface area of pile.
(Ulvi̇ , 2018), analyzed the utility of (Siebert & Teizer, 2014), assessed the
unmanned aerial vehicle in volume calculation. performance of UAV system for earthwork
The volume calculation was made by both calculation and compared to the conventional
photogrammetric technique and conventional survey techniques. The process of mapping the
survey methods. Ground Control Points (GCPs) study area was performed using quadrocopter
were installed for geo-referencing and evenly UAV at height of 50 m above the ground with
distributed throughout their area of study, in 70% and 40% forward and side overlap
addition to 1415 points were measured with respectively. The obtained images were
spaces of 40 cm to calculate the volume by using processed and point clouds generated by utilizing
Topcon GPT 8203A total station reflector-less. Agisoft Photoscan Professional. In classical
The process of taking photographs was carried survey GNSS receiver and SAPOS (German
out using Octocopter UAV at 20 m high. The DGNSS Reference Station System) was used for
captured images were processed with measuring 8 ground control points GCPs for geo-
Photomodeler software. Finally, the volume referencing and hundred single points using
comparison was made by both mentioned traditional RTK- GPS technique. The volumes of
methods, and the results showed that, volume three earth piles were compared by both methods,
acquired by photogrammetric technique agreed and the results agreed with each other within 59
with the volume got by classical method by m3(8%) for Pile 1, 93 m3(9%) for Pile 2, and 14
99.33%. m3(16%) for Pile 3.
(Akgul et al., 2017), conducted a research on (Tucci et al., 2019), UAV system was used for
the assessment of DEMs derived from both UAV monitoring and computing stockpile volumes of
and GNSS for earthwork volume. (Trimble UX5) bulk materials. The ground control points GCPs
drone was used in the process of study area for geo-referencing were obtained from the 3D
mapping with 80% longitudinal and traversal point model which was generated with terrestrial
coverage at flying height 200 m. All image data laser scanning. The laser scan data were
were processed by means of Agisoft Photoscan processed with Leica Cyclone software. In UAV
Professional software in order to produce point survey, DJI Phatom4Pro multirotor quadcopter
cloud and orthomosaic data. GNSS-based DEM was used for the process of image capturing with
was generated by measuring 5478 points with the two flights. The flight altitude was 75 m above
help of Network Real Time Kinematic (NRTK), the ground with 85% longitudinal and 70%
GNSS receiver and Pentax SMT-888 depending traversal coverage where the positions of the
on NRTKs. Turkey Continuously Operating cameras were vertically situated for the first one
Reference Stations (CORS). At the end, the result and the other obliquely situated at a (30˚) angle.
showed that the volume of excavation and All captured images were processed with the
embankment obtained by both techniques were software Agisoft Photoscan. The vertical position
very close to each other. was considered as a base reference and the
(Cryderman, Mah, & Shufletoski, 2014), volume of several stockpiles were computed with
evaluated the accuracy of earthwork computation both ArcGIS and Agisoft Photoscan. The results
of UAV photogrammetry against GNSS on a indicated that the volumes were well-matched
stockpile. In photogrammetric technique, in- with each other in a ratio of 99%.
house UAV was used to be flown multiple times (Ahmad, Dutsenwai, Periola, & Falowo,
over the study area with average height of 118 m 2017), analyzed volume computation of a water
and 75% frontal and side overlap. The acquired tank using low-cost close range photogrammetry.
image data were processed by means of Agisoft Phantom 3 Professional UAV was utilized to take
PhotoScan Professional (version 1.0.4). In the images for making 3D model of the water tank at
other method, Trimble R8s receiver was used to low altitude of 20 m from the ground surface. The
measure 11 ground control points GCPs for geo- UAV-based data were processed by Agisoft
referencing and 220 detail points using GNSS- Photoscan software. Their study, focus on the
RTK technique. Finally, the volume comparison influence of the number of images and ground

13
 control points GCPs were tested to select the between them not exceed 15m. These points were
optimum number of images and GCPs to get the depended on GNSS and digital elevation model
actual volume of the water tank. The 3D was generated.
coordinates of the GCPs were acquired from the On the other hand, Octocopter (8 rotors) UAV
traversing technique. The results showed that was used in the process of taking images at flight
using four GCPs and 95 images can get the water heights of 60m and 100m. After processing the
tank volume within 5% error of the actual collected data, dense point clouds, orthophoto
volume, as well as it was also demonstrated that mosaics and surface models were generated. The
the use of 4 GCPs and (115 to 220) images were two obtained surface model were compared and
needed to get the actual volume of the water tank. results analyzed. Finally, the results showed that
(Chunsen & Qiyuan, 2018), computed the the difference between volumes acquired by both
volume of coal pile using UAV images and methods did not exceed 1%.
compared to the results obtained by conventional (Arango & Morales, 2015), made a
RTK survey. DJI M600 six-rotor drone was used comparison of calculating stockpile volumes
to take images of the study area with 60% between UAV and Total Station with the actual
Heading overlap and 50% Lateral overlap. The volume. Two techniques were used for data
acquired photos were processed based on the acquisition, the first one with a total station, and
SfM-PMVS. Finally, the volumes of pile coal the other with a multicopter UAV. Leica TS02
obtained by both methods were compared to each plus, total station with 1" angular accuracy
other, and it found that, the UAV-based volume Reflector-less was used to compute volume of the
was 30374.1 m3, and the volume calculated using stockpile. Ground control points GCPs were
classical method was 30446.3 m3. The two measured with GNSS around the pile for geo-
volumes calculated agreed with each other within referencing. DJI Phantom 2 vision plus was used
72 m3 or 0.238 % error of the total volume. to map the test site at altitude of 50m above the
(Samad, Nekmat, & Rg, 2018), evaluated the ground surface, and Pix4D Mapper software was
DTM model generated by using UAV sensors in used to process the captured images. The volumes
earthwork calculation at quarry area. The UAV obtained by UAV and Total Station were
flight carried out using fixed-wing eBee UAV at compared to the actual volume. It was found that
altitude of 325m from the ground surface with 75 there was difference between UAV-based volume
% frontal and lateral overlap. The collected image and actual volume by -0.67% error, while 2.88%
data were processed with Agisoft Photoscan error between volume acquired by Total Station
software. The volume comparison analysis was and the actual volume.
made between UAV with and without GCPs (Stalin & Geoinformatics, 2017a), made a
depending on their contour interval. The six comparison between volume computation carried
GCPs were established by GPS using RTK out by both GPS and the UAV in an open pit
technique. Contours ranging from 1m to 10 m quarry. Leica viva GS08 plus was used to
interval generated relying on the data from Origin measure 7 GCPS equally distributed over the
Surveying Services Company. In conclusion, the study site and other detail points, the volume was
results demonstrated that the suitability of UAV- calculated using AutoCAD Civil 3D. On the other
based DTM for volume computation at quarry hand, eBee fixed-wing was used in the process of
area. mapping the study area at 118 m altitude above
(Julge, Ellmann, & Köök, 2019) used the ground surface with 75% frontal and lateral
unmanned aerial vehicle for monitoring overlap of the images. All gained images were
earthwork of road construction. The authors processed and volume computed with Pix4D
investigate the effect of different heights and the Mapper software. The results illustrated that an
numbers of Ground Control Points were agreement within 98.9% of the total volume
analyzed. Thirteen ground control points GCPs between both methods.
were signaled and equally distributed over the
study area using GNSS-RTK technic. Receiver of
GPS/ GLONASS Trimble R8 and Virtual
Reference Station (VRS) service provided via
data-link by a commercial CORS network were
used to measure 196 detail points with distance

14
 3. METHOD AND MATERIALS mountainous with size of approximately 80 m
wide and 573 m long and an average elevation of
3.1. STUDY AREA 615.5 m above mean sea level. The body of the
The test site is located at Duhok Dam in the Duhok Dam was taken as an application site due
city of Duhok in Kurdistan-Iraq. The boundary of to the fact that it was bare and with no vegetation
the study area lies within 36°52'33.41"N and 43° such as grass, shrubs and etc.
0'15.34"E (Fig. 2). The place was partially

DuhokDam
Duhok Dam

Water level
Water Level

Fig/(2): Location map of Study Area

3.2. UAV-SURVEY lines for each one cover 7, 4, and 3 lines

Rotary wing UAV platform (Phantom 4 PRO) respectively as illustrated in Table 1. frontal and
was used for mapping of the survey area as shown lateral overlap were both set to 80%, Table 1.
in (Fig. 3). UAV platform was equipped with a 1- In the field study, total 11 ground control
inch CMOS (Complementary Metal Oxide points GCPs were signalized using traditional
Semiconductor) 20-megapixel sensor and 8.8 mm GPS-RTK mode to assist spatial referencing
focal length, and has a manually adjustable process and geo-tagging of images which implies
aperture from F2.8 to F11. It also supports auto to compute the scale, orientation and absolute
focus by half-pressing the shutter button, and has position of the outputs in a desired coordinate
a focus range, from 1m to infinity as well as five- system as shown in Figure 5. Control points were
directions sensors for avoiding obstacle (DJI, measured with the help of Leica viva GNSS
2018). The aperture adjusted automatically to GS10, base and Leica viva GNSS GS15, rover
achieve the desired shutter speed. Three flights considering UTM WGS84- 38N coordinate
missions were flown over the test site at different system. Ground control points (GCPs) were
altitudes of 25 m, 50 m and 100 m, the numbers designed by 0.6 m× 0.6 m in size, and printed on
of captured images for each flight height were the piece of Flex as shown in Figure 4 which is a
684, 242 and 100 images respectively. The flight thin, plastic material coated a fabric at its back

15
 giving it extra strength to resist severe weather images taken at high altitude, and their centers to
conditions such as heat, rain etc. Solvent machine be reliably identified as well.
is used to print this Flex with a special ink that Pix4D Capture was used for flight planning
does not wear out because of heavy rain or which is an autopilot application available on the
sunlight. All GCPs were evenly distributed in market for both systems Android and IOS. The
somehow that cover the application area as shown software allows the user to select the desired
in figure 5. The shape of the GCPs was checker height of flight, percent of overlap of images and
pattern with a black and white square so as to shape of the mapping area. After powering on and
provide good contrast and ensure visibility in the pressing single button of START, then it will fly
the planed mission automatically (Pix4D, 2019).

Table (1): Flight planning parameters.

Flight altitude (m) 25 50 100

Flight time (min) 26 10 5

Surface area (m2) 45840

Forward overlap % 80%

Side overlap % 80%

GSD cm/pixel 0.68 1.36 2.73

Flight lines 7 4 3

No. of images 684 242 100

GPS propellers

Motor

Vision system

Gimbal and Camera

Fig. (3): Phantom 4PRO.

16
Proposed Target

Proposed Target

Fig. (4): Sample of the Ground Control Points (GCP)s

A3
A2
A1

GCPs
Survey Points

Fig. (5): Study area used for volume calculation with survey points and GCPs.

3.3. GPS- RTK SURVEY respect to the accuracy, stated by the

In the recent years, the GNSS technologies manufacturer ±15mm ± 10 ppm RMS horizontal
have become an important part of the geodetic and ±20mm ± 10 ppm vertical. The GPS-RTK
world. There are various types’ receivers for mode was used for measuring all GCPs
different purposes. Generally, RTK receivers are coordinates as shown in figure 6 and about 200
used for engineering tasks (Raeva, Filipova, & volume survey points along the crest of the
Filipov, 2016). Leica viva GNSS was used in this Duhok Dam including all characteristic points
study, GS10 as a base and GS15 as a rover as those create the terrain model such as edges as
shown in Figure 6. The following technical shown in Figure 5. The space between points did
parameters of GNSS, GS10 and GS15 with not exceed 25 m. The base was set up on a

17
 benchmark point with accurate and adjusted three GmbH)(Devers, 2018).the point data were
dimensional coordinates (X, Y, and Z) obtained imported into AutoCAD Civil3D for volume
from manned Aerial Survey carried out by a calculation.
Germany company named (GEOCART

Antenna

GS10 Rover

Survey Pole
GS10 Base
Controller

Point centered on
the proposed target

Fig. (6): Leica Viva GS10, base and Leica viva GS15, rover.

3.4. Methods of Volume Computation one of them is the natural surface (upper surface)
There are four most frequently used methods and the second one is the reference surface or base
for volume computation for different purposes surface or design surface (lower surface); the
which are :(Tucci et al., 2019)(Julge et al., 2019): quantity enclosed by these two surfaces is
1. Cross sectional method. discretized into elementary prisms with a
2. Grid method. triangular base as shown in Figure 7 or
3. Horizontal section (contour) method. trapezoidal prisms from which volume is
4. Prisms method. computed using the following formula 1 (case of
The prisms method between two surfaces is a triangular prism):(Tucci et al., 2019)
discussed in this paper which is the fundamental
source of the calculation used with the software. ∑3𝑖=1 𝐻𝑖
The prisms method comprises of the definition 𝑉𝑖 = 𝑃 1
3
of two surfaces, determined by using of
interpolation of three-dimensional spatial data,

18
 Fig. (7): Elementary triangular prism.

3.5. Data Processing process of volume computation was performed by

The GPS data collected using RTK technique comparing the DEM created using the GPS
were imported into the AutoCAD Civil 3D, and observations and the model that obtained by UAV
the elevation model was generated from these process (different reference surface).
three dimensional points above MSL. Next, the

Align Photos GPS

Measurements
Build Dense Build Mesh Build DEM Build
Point Cloud Orthomosaic

1. Camera
Image
Optimization.
Matching 2. Bundle
blocks
adjustment.
Fig.(8): Agisoft Software Workflow.

On the other hand, photogrammetric data were successive of overlapping photographs with the
processed using Agisoft Photoscan Professional corresponding referencing information (Agisoft,
software (version 1.4.3) founded in 2006 which 2017).
uses the Structure-from-Motion (SfM) that The software work flow begins with new
provides the users an opportunity of low-cost project creation, recommended setting
three-dimensional data acquisition (Akgul et al., configuration and importing images with their
2017)(Micheletti, Chandler, & Lane, 2015). metadata (camera details and settings).
Agisoft PhotoScan Professional permits the Then, the image quality function was
generation of geo-referenced dense point clouds, performed in order to examine which photo is
textured polygonal models, digital elevation blurred or distorted that may affect the results.
models and orthomosaics from a group of This is an algorithm used to analyze contrast

19
 between pixels, the higher the contrast, the dense point cloud DEM and Orthomosaic was
sharper the images (Agisoft LLC, 2013). Poor generated as a final photogrammetric product
input, e. g. ambiguous photos can influence (Fig.12).
alignment results badly. This feature helped to As the DEM is built, it is possible to measure
exclude poorly focused images from point, distance, area and volume. The software
photogrammetric processing.by removing images allows the volume measurement above three
with quality value of less than 0.5 units, providing planes (reference surface) which are best fit,
that the rest of the photos cover the whole scene mean level and custom level planes. Best fit plane
to be reconstructed. This function has a scale is an inclined surface calculated using
ranged from 0 to 1, under 0.5 considered distorted interpolation of vertices of polygon drawn. Mean
or unwanted image (Agisoft, 2019). Fortunately, level plane is horizontal flat surface calculated
all images were accepted with a scale more than using medium height determined by the heights
0.8. The main processing steps as shown in Figure of the vertices of polygon drawn. The third option
8 were executed according to the recommended is custom level plane which is horizontal flat
parameters setting by the Agisoft manual. surface at a reference height defined by the user.
The first step which is aligning photos was The last option was used in this study is the
lunched. At this phase, the software carries out elevation levels of (596 to 610 m) at 1m interval.
the point matching process among the overlapped In addition to three polygons or boundaries with
images and obtains the rough camera position and different areas, were imported into the software as
orientations for each photo and builds cloud a shape files as shown in Figure 5 which were the
model based on tie point. The next step was same boundaries used by the AutoCAD Civil 3D
importing markers or GCPs into the software with for volume computation.
txt supported file for the purpose of geo-
referencing and optimization of camera positions 4. RESULTS AND DISCUSSION
and orientation data. The point cloud was 4.1. Results
transformed to WGS84-38N coordinate system
automatically by the software. Furthermore, Table 4 demonstrates the calculation of
depending on the estimated camera position, the volumes obtained by two different techniques
software computes the depth information for each photogrammetric and GPS. In photogrammetric
camera in order to be combined into a single method, the volume obtained at three different
dense point cloud. The medium quality was altitudes (25,50 and 100) m with the variation of
chosen for dense point cloud building, higher elevations of reference surface. The differences
quality takes longer time. After that, build mesh between volumes acquired by both techniques are
was built by selecting the height field option for calculated in cubic meters and in percentages.
the surface type (Samad et al., 2018). Based on

Table (2): calculation of volumes obtained by two different techniques GPS and photogrammetric at
three different altitudes (25,50 and 100) m.

Area3=17322.89 m2
3
Elevatio Photogrammetric volume m GPS Difference m3 Differences %
nm/ Volume
Altitude m3
m A(25m) B(50m) C(100m) D A-D B-D C-D (A- (B- (C-
D/A)*10 D/B)*10 D)/C)*10
0 0 0
610 110186.1 110291.1 110316.1 110030.6 155.5 260.4 285.4 0.14 0.24 0.26
8 0 0 4 4 6 6
609 127508.6 127614.4 127639.4 127353.5 155.1 260.8 285.8 0.12 0.20 0.22
4 0 0 2 2 8 8
608 144831.1 144937.7 144962.7 144676.4 154.7 261.2 286.2 0.11 0.18 0.20
2 0 0 1 1 9 9

20
 607 162153.7 162261.0 162286.0 161999.3 154.4 261.7 286.7 0.10 0.16 0.18
0 0 0 0 0 0
606 179476.1 179584.3 179609.3 179322.1 153.9 262.1 287.1 0.09 0.15 0.16
8 0 0 9 9 1 1
605 196798.6 196907.6 196932.6 196645.0 153.5 262.5 287.5 0.08 0.13 0.15
6 0 0 8 8 2 2
604 214121.1 214230.9 214255.9 213967.9 153.1 262.9 287.9 0.07 0.12 0.13
4 0 0 7 7 3 3
603 231443.7 231554.2 231579.2 231290.8 152.8 263.3 288.3 0.07 0.11 0.12
0 0 0 5 5 5 5
602 248766.1 248877.5 248902.5 248613.7 152.4 263.7 288.7 0.06 0.11 0.12
8 0 0 4 4 6 6
601 266088.6 266200.8 266225.8 265936.6 152.0 264.1 289.1 0.06 0.10 0.11
6 0 0 3 3 7 7
600 283411.1 283524.1 283549.0 283259.5 151.6 264.5 289.4 0.05 0.09 0.10
4 0 0 2 2 8 8
599 300733.7 300847.4 300872.3 300582.4 151.3 264.9 289.8 0.05 0.09 0.10
2 0 0 1 1 9 9
598 318056.1 318170.7 318195.6 317905.2 150.8 265.4 290.3 0.05 0.08 0.09
8 0 0 9 9 1 1
597 335378.6 335494.0 335518.9 335228.1 150.4 265.8 290.7 0.04 0.08 0.09
6 0 0 8 8 2 2
596 352701.1 352817.3 352842.2 352551.0 150.0 266.2 291.1 0.04 0.08 0.08
4 0 0 7 7 3 3

Figure 9 and 10 illustrates the relationship between the elevations of reference surface and
differences in volumes between photogrammetric method and GPS technique.

Volume Differences
0.30

0.25

0.20
Percents %

0.15

0.10

0.05

0.00
610 609 608 607 606 605 604 603 602 601 600 599 598 597 596
Reference surface Elevations m

25 m 50 m 100 m

Fig.(9): Relationships of Reference surface levels with differences in percent at (25 m,50 m and 100
m) altitudes.

21
 Volume Differences
0.30

0.25

0.20
PERCENTS %

0.15

0.10

0.05

0.00
610 609 608 607 606 605 604 603 602 601 600 599 598 597 596
REFERENCE SURFACE ELEVATIONS M

25 m 50 m 100 m

Fig (10): comparison between Volume differences in percent and Reference surface levels.

25m Height 50m Height

Crest can be shown clearly Crest can be shown

100m Height

Crest can be estimated

Fig. (11): Digital Elevation Models DEMs for 25m, 50m and 100m altitudes.

4.2. Discussion different data acquisition techniques which were

In this paper, digital elevation model DEM UAV and GPS. an area of (17322.89 m2) was
generated by means of UAV images was assessed selected (Fig.5) with different elevations of
by comparing the volumes obtained by two reference surface ranged from 596m to 610m in

22
one-meter interval. As the elevation of reference traditional method with total station as well as it
surface is increased, the difference in volumes is faster than total station by six times in term of
increased as shown in figure (9 and 10). data acquisition.(Wang et al., 2017), also proved
the suitability and preference of UAV in
The accuracy of volume calculation depends computing actual earthwork amounts compared
on many factors such as flight height, surface- to the traditional GPS method with an error
volume ratio, surveying method, DEM, type of ranged from 0.1% to 0.9% of the total earthwork
software used for processing and etc. Therefore, quantities.
the final volume quantities are determined in 5. CONCLUSION
relative terms not absolute terms. Surface-volume In this study, the accuracy of the Unmanned
ratio is one of the factors that taken into account Aerial Vehicle UAV photogrammetry for volume
in this study, the smaller the surface volume ratio, computation comparing to the traditional GPS
the more accurate the volume as shown in (table method was analyzed. The results showed that,
2). This is also explained by (Tucci et al., 2019), the volume calculated using UAV-based DEM
as the surface-volume ratio decreases, the agreed with the volume obtained by GPS-RTK
computed volume will be more accurate. The technique in the ratio of 99.74% (0.26%
impact of flying height is also considered, the error),99.76 and 99.86 for flight heights (100,50
higher flight provides larger area coverage, but and 25) m respectively as shown in table 2. The
with larger value of ground sample distance GSD worst ratio that is acceptable according to the
too. However, the higher flight altitudes affect the latest literature ranges from 0.1% to 5% of the
value of GSD and level of details of the model, total volume. Flight heights lower than 100 m
but it doesn’t create systematic shifts (Julge et al., improve the general accuracy of the results, and
2019).The lower the flight height, the more provide a better achievement of some finer
accurate the volume. The maximum difference surface details too.
occurred in the case of flight height 100m which
was 0.26% difference (99.74% agreement)
between volumes calculated by both methods, or REFERENCES
4.4 cm thickness over the entire surface area. In Agisoft. (2017). Tutorial (Beginner level):
contrast, minimum difference was 0.14 % in the Orthomosaic and DEM Generation with
case of 25m altitude when elevation of base Agisoft PhotoScan Pro 1.2 (without Ground
surface was 610 m. Generally, the best flight
Control Points). 2, 1–14.
height among the three flight altitudes was 25m,
https://doi.org/10.1006/mpev.1996.0003
while there was no significance difference
between 50m and 100m altitudes in volume Agisoft. (2019). Agisoft PhotoScan. Retrieved
computation. October 22, 2019, from Agisoft Metashape
website: https://www.agisoft.com/
Accuracy of DEM is an important factor for Agisoft LLC. (2013). Agisoft PhotoScan Professional
volume computation. Previous studies indicated Edition User Manual, Version 0.9.1. 62.
that it is possible UAV images for estimating Retrieved from
volumes with sufficient accuracy (Stalin & http://www.agisoft.com/features/professional-
Geoinformatics, 2017). Accurate DEM generated edition/
using UAV images provides cost effective data
Ahmad, B., Dutsenwai, S., Periola, A. A., & Falowo,
and proved that it is time-saving for earthwork in
engineering works especially in road designing O. E. (2017). Volumetric calculation using low
on lands with low vegetation or bare ground cost unmanned aerial vehicle ( UAV ) approach
(Akgul et al., 2017).(Ulvi̇ , 2018) accentuated that, Volumetric calculation using low cost
there was a fourfold increase in the time required unmanned aerial vehicle ( UAV ) approach.
for volume calculation using classical methods https://doi.org/10.1088/1757-
than using UAV techniques as well as there was 899X/270/1/012032
99.33% agreement between volumes calculated Akgul, M., Yurtseven, H., Gulci, S., & Akay, A. E.
by UAV and GPS methods. (Arango & Morales, (2017). Evaluation of UAV- and GNSS-Based
2015), demonstrated that volume estimation by
DEMs for Earthwork Volume. Arabian Journal
means of UAV-based DEM is more accurate than

23
 for Science and Engineering. STOCKPILE – A STUDY CASE COMPARING
https://doi.org/10.1007/s13369-017-2811-9 GPS AND UAV MEASUREMENTS IN AN
Arango, C., & Morales, C. A. (2015). Comparison OPEN PIT QUARRY. C(February 2019).
between multicopter UAV and total station for https://doi.org/10.5194/isprsarchives-XLI-B1-
estimating stockpile volumes. International 999-2016
Archives of the Photogrammetry, Remote Samad, A. M., Nekmat, M. S., & Rg, A. (2018).
Sensing and Spatial Information Sciences - Assessment of Generated DTM Model Using
ISPRS Archives, 40(1W4), 131–135. UAV Sensors Toward Earthwork Calculation.
https://doi.org/10.5194/isprsarchives-XL-1- 2018 IEEE 8th International Conference on
W4-131-2015 System Engineering and Technology (ICSET),
Chunsen, Z., & Qiyuan, Z. (2018). Research on (October), 40–45.
Volumetric Calculation of Multi-Vision Siebert, S., & Teizer, J. (2014). Mobile 3D mapping
Geometry UAV Image Volume. 2018 Fifth for surveying earthwork projects using an
International Workshop on Earth Observation Unmanned Aerial Vehicle (UAV) system.
and Remote Sensing Applications (EORSA), 1– Automation in Construction, 41, 1–14.
4. https://doi.org/10.1016/j.autcon.2014.01.004
Cryderman, C., Mah, S. B., & Shufletoski, A. (2014). Stalin, J. L., & Geoinformatics, M. S. (2017). Volume
EARTHWORKS COMPUTATIONS. 68(4), Calculation from UAV based DEM. 6(06), 126–
309–317. 129.
Devers, W. (2018). A Case Study on the Use of Drones Tucci, G., Gebbia, A., Conti, A., Fiorini, L., &
on Heavy Civil Construction Projects. Lubello, C. (2019). Monitoring and
DJI. (2018). Flight Controller Data Analysis Series Computation of the Volumes of Stockpiles of
Tutorials--Part 1 (p. 1). p. 1. Bulk Material by Means of UAV
Julge, K., Ellmann, A., & Köök, R. (2019). Photogrammetric Surveying.
UNMANNED AERIAL VEHICLE SURVEYING Ulvi̇ , A. (2018). Analysis of the Utility of the
FOR MONITORING ROAD. 14(1), 1–17. Unmanned Aerial Vehicle (Uav) in Volume
Micheletti, N., Chandler, J. H., & Lane, S. N. (2015). Calculation By Using Photogrammetric
Structure from Motion (SfM) Photogrammetry. Techniques. International Journal of
In British Society for Geomorphology Engineering and Geosciences, c, 43–49.
Geomorphological Techniques (Vol. 2). https://doi.org/10.26833/ijeg.377080
Retrieved from www.photosynth.net Wang, X., Al-Shabbani, Z., Sturgill, R., Kirk, A., &
Pix4D. (2019). Pix4D. Retrieved October 22, 2019, Dadi, G. B. (2017). Estimating earthwork
from Pix4D website: https://www.pix4d.com/ volumes through use of unmanned aerial
Raeva, P., Filipova, S. L., & Filipov, D. G. (2016). systems. Transportation Research Record,
VOLUME COMPUTATION OF A 2630, 1–8. https://doi.org/10.3141/2630-01
STOCKPILE – A STUDY CASE COMPARING
GPS VOLUME COMPUTATION OF A

24