International Journal of Remote Sensing

ISSN: 0143-1161 (Print) 1366-5901 (Online) Journal homepage: https://www.tandfonline.com/loi/tres20

Is UAV-SfM surveying ready to replace traditional

surveying techniques?

J. J. Carrera-Hernández, G. Levresse & P. Lacan

To cite this article: J. J. Carrera-Hernández, G. Levresse & P. Lacan (2020) Is UAV-SfM surveying
ready to replace traditional surveying techniques?, International Journal of Remote Sensing, 41:12,
4818-4835, DOI: 10.1080/01431161.2020.1727049

To link to this article: https://doi.org/10.1080/01431161.2020.1727049

Published online: 02 Mar 2020.

Submit your article to this journal

Article views: 20

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tres20
INTERNATIONAL JOURNAL OF REMOTE SENSING
2020, VOL. 41, NO. 12, 4818–4835
https://doi.org/10.1080/01431161.2020.1727049

Is UAV-SfM surveying ready to replace traditional surveying

techniques?
J. J. Carrera-Hernández , G. Levresse and P. Lacan
Centro De Geociencias, UNAM. Blvd, Querétaro, México

ABSTRACT ARTICLE HISTORY

The creation of high-resolution Digital Elevation Models (DEMs) Received 18 January 2019
through the acquisition of images with a consumer-grade camera Accepted 21 December 2019
mounted on a light Unmanned Aerial Vehicle (UAV) and their
processing with Structure from Motion (SfM) software has been
widely adopted in different fields. Despite this wide adoption, its
use has been limited to research projects – as in our experience
most practising surveyors, think that the topography obtained
through this technology is subpar to that obtained with surveys
undertaken with traditional methods. To address this concern, we
compare these two approaches in a heterogeneous built-up area of
nearly 33 ha with an elevation between 1899 and 1944 m a.s.l. This
comparison shows that the UAV-SfM methodology provides better
results than the traditional method, is less prone to human error
and also requires significantly less time: 4 h of fieldwork instead of 1
week. Furthermore, the spatial resolution of a Total Station (TS)
survey is only limited to those points surveyed, whereas the UAV-
SfM derived topography is a subproduct of the resultant three
dimensional point cloud, which can also be used to generate a
high-resolution orthophoto.

1. Introduction
The advances in surveying technology have improved the way in which the elevation of
land surface can be sampled: from ground surveying using theodolites and Total Stations
(TS) – a time consuming task that requires highly skilled and meticulous surveyors
(Nelson, Reuter, and Gessler 2009) – to aerial and satellite surveys that make use of
photogrammetry and more recently to airborne laser scanning known as Light
Detection and Ranging (LiDAR). The use of LiDAR – with both high sampling density
and vertical accuracy – has enabled the production of high-resolution Digital Elevation
Models (DEMs); unfortunately, the acquisition of LiDAR data is expensive. However, the
development of both high-resolution topography and orthophotos is now possible with
the use of Structure from Motion software (SfM) and consumer-grade cameras mounted
on light Unmanned Aerial Vehicles (UAVs). Structure from Motion is a computer vision
technique that involves the simultaneous recovery of 3D camera motion and 3D scene
structure from a collection of tracked 2D features on overlapping pictures (Szeliski 2011).

CONTACT J. J. Carrera-Hernández jaime-carrera@geociencias.unam.mx Centro De Geociencias, UNAM. Blvd,

Juriquilla 3001, Querétaro CP 76230, México
© 2020 Informa UK Limited, trading as Taylor & Francis Group
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4819

By using images acquired with an UAV, a three-dimensional point cloud can be generated,
with a quality and resolution comparable – or even better – than a LiDAR-generated point
cloud (Leberl et al. 2010; Fonstad et al. 2013). Because this technology makes it possible to
generate high-resolution topography that was only achievable through more costly and
labour-intense methods, it has been widely adopted for different purposes: monitoring of
sand dunes and beaches (Gonçalves and Henriques 2015), analysis of glacier dynamics in
Greenland and the Himalayas (Immerzeel et al. 2014; Ryan et al. 2015), fluvial topography
(Woodget et al. 2015; Javernick, Brasington, and Caruso 2014), landslide mapping
(Niethammer et al. 2012), soil erosion (D’Oleire-Oltmanns et al. 2012; Castillo et al.
2012), mapping of both fractures and structures on a sinking bottom lake located in a
maar (Carrera-Hernández et al. 2016), mapping of fault zone topography (Johnson et al.
2014) and even to study morphological and structural changes on a lava dome
(Darmawan et al. 2018). Other studies have used the SfM point clouds to inspect dams
(Buffi et al. 2017), to determine stockpile volumes (Arango and Morales 2015), to map
fractures in a marble quarry (Salvini et al. 2017) and even for morphological analyses of
buildings (Teza, Pesci, and Ninfo 2016). The aforementioned studies have used either
helicopters, motorized gliders, helium balloons, or UAVs (fixed-wing and multirotor, as
detailed in Carrera-Hernández et al. (2016)).
Various studies have analysed the accuracy of the UAV-SfM approach; some of these
studies have focused on which of the existing Structure from Motion software (Agisoft
Photoscan, MicMac, Pix4UAV, and Blunder) yield better results (Ouédraogo et al. 2014; Sona
et al. 2014; Turner, Lucieer, and Wallace 2014), while others have assessed the accuracy of the
Digital Elevation Models generated with SfM, using either LiDAR or Terrestrial Laser Scanner
(TLS) point clouds as benchmarks (Westoby et al. 2012; Ouédraogo et al. 2014). Ouédraogo et
al. (2014) compared the Digital Terrain Model obtained with a TLS and the one derived with
two software packages that use Structure from Motion: MicMac and Photoscan. Westoby et al.
(2012) compared a Structure from Motion derived DSM with one obtained by using a TLS on
an exposed rocky coastal cliff, a breached moraine-dam complex and a glacially sculpted
bedrock ridge; they found that even decimetre-scale vertical accuracy can be achieved with
Structure from Motion. Although surveying with a TLS has benefits, its disadvantages are that
multiple scan positions are likely to be necessary – to both avoid blind spots and gain spatial
coverage – in addition to its high cost and the requirement to locate and georeference targets,
which have to be achieved with either a TS or a differential Global Positioning System (GPS,
Carrivick, Smith, and Quincey (2016)).
Georeferencing of the SfM point cloud is required to translate it to a given coordinate
system (i.e. Universal Transverse Mercator (UTM)); this is a task that is generally achieved
through the use of Ground Control Points (GCPs) and can be avoided by direct georeferen-
cing (Turner, Lucieer, and Wallace 2014), which consists of synchronizing a GPS located on
the UAV with its camera shutter and writing its coordinates on each picture’s Exchangable
Image File format (EXIF) file. This is particularly useful now that Real-Time Kinematic (RTK)
GPS technology is available on some drones. Although direct georeferencing avoids the use
of GCPs to georeference the resulting point cloud, they are still required to validate the
resulting DEM. Accordingly, one question that has been addressed in different works is the
effect of the quantity and spatial distribution of GCPs on the generated point cloud (Tonkin
and Midgley 2016; Agüera-Vega, Carvajal-Ramírez, and Martínez-Carricondo 2017; Gindraux,
Boesch, and Farinotti 2017; James et al. 2017). According to James et al. (2017) the number
4820 J. J. CARRERA-HERNÁNDEZ ET AL.

of CGPs to be used depends on the area to be surveyed and on the equipment used;
therefore, the use of cameras with high levels of lens distortion has to be avoided because
the resulting point cloud will be susceptible to much greater error.
Total Stations (TS) are the predominant instruments used in surveying practice today
(Wolf 2003; Ghilani 2017) due to their capacity of measuring both angles and distances in
addition to automatically display the results and store them. They have been used in
different studies and areas, such as lanslide monitoring (Artese and Perrelli 2018), measure-
ment of streambank erosion (Myers, Rediske, and McNair 2019), to obtain bathymetric data
(Parsapour-Moghaddam and Rennie 2018) to determine elevations of hydrometric gauge
stations (Baiocchi et al. 2018), to map archeaological sites (Venter, Shields, and Cuevas
Ordóñez 2018) and to locate vegetation quadrats (Dodd 2011). In practice, most surveying
projects are still undertaken with Total Stations (TS), because TS are considered to yield the
most accurate measurements (Nelson, Reuter, and Gessler 2009). However, the accuracy
achieved with a TS is related to operator procedures as well as the condition of both the TS
instrument and other peripheral equipment used with it (Ghilani 2017).
Some studies that have compared the results between surveys undertaken with a TS
and SfM are those of Arango and Morales (2015) who compared traditional surveying
methods to those obtained with Structure from Motion to estimate stockpile volumes or
Tonkin et al. (2014) who compared the results of the TS survey reported by Graham and
Midgley (2000) with a DEM developed with the UAV-SfM approach by determining the
Root Mean Square Error (RMSE) as accuracy measure. The TS survey consisted of 7,761
points, which were used to validate the Digital Elevation Model, obtaining a general
RMSEv = 0.517 m with RMSEv = 0.200 m in sparsely vegetated areas. It should be kept in
mind that the TS survey of Graham and Midgley (2000) was developed to construct a
high-resolution DEM on approximately 9 ha. The comparisons that have been undertaken
between traditional surveying techniques and SfM are generally restricted to small areas
and mainly for research purposes. This work aims to assess if there are any advantages of
using the UAV-SfM approach (in both time and sampling density) over a survey under-
taken with a TS, which is – in general – the procedure followed by practising surveyors. To
achieve this goal, we surveyed an area of nearly 33 ha with the UAV-SfM approach and
compared our results with a TS survey undertaken by a contractor on the same area.

2. Study area
The study area is a campus of Mexico’s National University (UNAM), located in Queretaro –
approximately 200 km north of Mexico City – and where our research group is based. The
advantage of this study area is that it is located on a hill with a gentle slope – enabling us
to measure a smooth topography change – has controlled access, and comprises both
paved and pedestrian roads, buildings and vegetated areas, which provide both smooth
and drastic elevation differences. The area covered by this campus is 32.767 ha
(0.327 km2), with an elevation difference of 55 m (with minimum and maximum values
of 1899 and 1944 m a.s.l., respectively). Because this campus is located in a semiarid
region, shrubs are the main vegetation type with some cacti also present; some of these
areas are densely vegetated, making it impossible to walk through them without dama-
ging the existing vegetation.
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4821

3. Methodology
3.1. UAV setup
For this work, a Phantom 2 quadcopter (manufactured by DJI Innovations: http://www.dji.
com/product/phantom–2) was used. Without propellers, the diagonal length of this quad-
copter is 39 cm; with propellers, this length increases by 20 cm and can carry a point and
shoot camera in addition to its battery, which has a capacity of 5.2 Ah, allowing the
quadcopter to fly for about 13 min when carrying the setup used in this work. This setup
included: (a) a GPS watch (Garmin Forerunner 305) attached to the upper part of the
quadcopter, (b) a Canon S110 camera pointing downwards and attached on the lower
part of the quadcopter’s body, and (c) DJI’s 2.4 GHz datalink (installed inside the quadcop-
ter’s body) that allows communication between a PC and the quadcopter for flight pro-
gramming. The Phantom 2 quadcopter has a GPS enabled flight controller (Naza-M V2) with
accuracies of 2.5 m on the horizontal and 0.8 m on the vertical. The GPS watch is used to
record flight details (i.e. coordinates, elevation, and speed) and after all flights are completed
a .gtx file is extracted from the portable GPS. These extracted coordinates are then written
on the picture’s EXIF header by matching the acquisition time of each picture and the
coordinates recorded by the GPS at the acquisition time through the use of Digikam, a free
and open-source image organizer and editor. The advantages of writing the coordinates of
each image on its EXIF file are faster processing times in PhotoScan and direct georeferen-
cing of the resulting point cloud. This approach is similar to that of Girod et al. (2017) who
used a GoPro attached to a helicopter and a Garmin GPSmap to register the flight’s
coordinates to monitor a glacier in Norway.
The Canon S110 camera was selected because it can be set in intervalometer mode by
using the Canon Hacker’s Development Kit (CHDK). The CHDK is written on the Secure
Digital (SD) card of the camera and is automatically loaded when the camera is turned on,
allowing extra features not available on the camera’s default firmware. Through the use of
the CHDK, the camera was set to take pictures every 2 s in raw format, focused at infinity,
International Organization of Standardization (ISO) 500, a shutter speed of 1 600 s, a focal
length of 11 mm (equivalent to 50 mm in a full-frame camera, due to the S110’s crop
factor of 4.62) and an aperture of f 8 (the maximum possible on this camera); the SD cards
used on the camera had a 32 GB capacity.
In order to cover the study area, a total of three flights were required (Figure 1(a)). These
flights were programmed to take place at a height above 100 m from the launching point and
for the images to have a lateral overlap of 60% – as required by Photoscan. Each flight covered
a length of nearly 4 km and at a real height of approximately 110 m above the ground – as
registered by the portable GPS (Figure 1(b)). Before flying the UAV above the residential area
located west of campus (Figure 1(a)) we talked with the residents and explained them why we
were flying the UAV, as UAVs have to be flown according to local regulations.
The camera was set-up to acquire a picture every 2 s – which satisfies Photoscan
requirements of 80% overlap on the forward direction while the drone flew at a velocity of
approximately 10 m s 1 . With this set-up, a total of 845 images were acquired and processed
with Agisoft Photoscan. Before importing the images into Photoscan, their EXIF file header
was modified in order to include the coordinates at which the multicopter was flying when
each image was acquired and to modify the default focal length written on the EXIF header of
4822 J. J. CARRERA-HERNÁNDEZ ET AL.

(a) (b)

Figure 1. Flights required to cover the study area: (a) plan view, (b) elevation profiles. The required
three flights were both launched and landed from the same point and covered approximately 12 km;
as can be seen from (b) the quadcopter flew at approximately 110 m above the terrain.

each image. As previously mentioned, the coordinates were written with the use of Digikam
through the use of the extracted .gtx file from the GPS watch while the focal length was
modified using the exiftools utility. To georeference the images, Digikam matches the
acquisition time of each image to the coordinates recorded by the GPS at acquisition time
– this requires the time on the camera to be synchronized with the GPS before flying the UAV.
Also, the focal length has to be written on the EXIF file because the camera we used had a
zoom-lens and focal length information is used by Photoscan to align the images.
Once the EXIF header of the 845 images was updated with both coordinates and focal
length, they were imported into Agisoft PhotoScan. After all required processing, a point
cloud of 38,964,391 points for the 32.767 ha (0.327 km2) was created, yielding a point
density of 119 points/m2 as shown in Figure 2. The resulting point cloud was able to
capture details such as windows in walls even though no oblique images were taken, as
this work was aimed at developing a Digital Elevation Model.

3.2. Validation of UAV-SfM survey

For this work, we did not place Ground Control Points (GCPs), but opted to use existing
features which were easily seen on the images. We first flew the study area and produced
a directly georeferenced orthophoto with Photoscan – without using GCPs – which was
printed at a large scale and used to locate the reference points. When each point was
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4823

Figure 2. Point cloud generated in Photoscan. The buildings and trees are adequately represented in
the point cloud, even though no oblique images were taken.

measured with the differential GPS, it was marked on the printed orthophoto and a
picture of the measured feature was taken, as shown on Figure 3. As can be seen on
this Figure, the points were located on different terrain features in order to capture the
variation in elevation (i.e. buildings or small walls). After the coordinates of the GCPs were
acquired in the field, they were imported and located in Photoscan in order to accurately
georeference the model and to determine its RMSE.
A total of 137 points distributed throughout the campus (Figure 4) were registered
with a Trimble R7 on Real-Time Kinematic mode with a horizontal precision of 0.08 m and
0.015 m on the vertical and set to only register coordinates with a maximum Position
Dilution of Precision (PDOP) of 4 (PDOP  4). These points were divided in two sets: one
was used to improve the georeference of the point cloud (n ¼ 62), while the other set was

Figure 3. Physical features used as both ground control points and validation points. These features
were chosen in order to measure the height variability in the study area and were thus located on
roofs, small walls, and flat terrain. Each point was labelled according to its GPS record.
4824 J. J. CARRERA-HERNÁNDEZ ET AL.

Figure 4. Spatial distribution of both ground control and validation points. The ground control points
were used to georeference all the acquired images in photoscan, while the validation points were
used as independent points to measure the spatial distribution of location errors. Coordinates are in
UTM-14.

used to validate the generated Digital Elevation Model (n ¼ 75). The first set of points
(n ¼ 62) was used in Photoscan to adequately georeference the image – because it was
originally referenced with the portable GPS – yielding a horizontal RMSE = 0.086 m and a
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4825

(a) (b)

Figure 5. Spatial distribution of errors for those points that were used as Ground Control Points in
Photoscan: (a) Horizontal, (b) Vertical. Coordinates are in UTM-14.

vertical RMSE = 0.076 m (Figure 5), with a maximum horizontal error of 0.34 m for one
point and an absolute error for two points on the vertical of 0.20 m (Figure 5).
To validate the topography obtained with the UAV-SfM approach, both the orthophoto
and the generated DEM were saved as GeoTiffs with a 0.10 m resolution and imported in
the Geographic Resources Analysis Support System (GRASS) Geographic Information
system (GIS, GRASS-Development-Team (2019)). A vector map was created by locating
the validation points on the orthophoto and their corresponding elevation on the DEM
was determined by querying the DEM with the location of the validation points through
the v.what.rast command in GRASS, yielding a RMSE = 0.129 m on the vertical. The
horizontal errors were determined by subtracting the x; y coordinates obtained with the
GPS from those obtained by locating the validation points on the orthophoto, yielding an
RMSE = 0.119 m (Figure 6).

3.3. Digital terrain model

The Digital Terrain Model (DTM) was obtained by removing the non-ground points with
the Progressive Morphological (PM) filter developed by Zhang et al. (2003) with further
filtering achieved with a statistical outlier removal method. In order to apply these
filters, the point cloud generated in PhotoScan was exported as a .las file, filtered
through the Point Data Abstraction Library (PDAL) library and visualized with Cloud
Compare (Figure 7). The PM filter creates a grid from the point cloud at a specified
resolution; for this work, a 1 m DTM was generated and then imported in the GRASS GIS
4826 J. J. CARRERA-HERNÁNDEZ ET AL.

(a) (b)

Figure 6. Spatial distribution of errors on validation points: (a) Horizontal, (b) Vertical. Coordinates are
in UTM-14.

for further processing, as the generated DTM with the PM filter still has null values where
large buildings were located (Figure 7). Once the DTM was imported into GRASS, a
contour vector map was generated using a 0.25 m interval; this contour map was then
used to interpolate the DTM in order for it to cover the entire study area with the v.surf.
rst command, which implements a regularized spline with tension interpolation
method, as described by Mitášová and Mitáš (1993). Finally, contour lines at 0.5 m
were extracted from this DTM and compared to those obtained with a Total Station
survey undertaken by an independent team of surveyors hired by the construction
department of the campus for another project – both surveys were undertaken with a
time difference of 1 month during the dry season. The contour lines developed with
these two technologies are shown in Figure 8, where their differences can be seen. It
should be mentioned that the time required to complete the fieldwork with the UAV-
SfM approach took approximately 4 h for a team of three people, while the independent
TS surveying team required 1 week.

4. Results and discussion

The contour lines shown in Figure 8 – although similar in some parts of the domain –
exhibit differences above 1 m in several regions. Evidently, this is a matter of under-
sampling by the surveyors who used the Total Station (TS), as can be seen on Figure 9,
which shows the spatial distribution of the 6,240 points used on the TS survey. To
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4827

Figure 7. Filtered point cloud showing only those points that correspond to terrain. The filtering was
done using photoscan’s classification algorithm and the progressive morphological filter with outlier
removal implemented in the point data abstraction library (PDAL).
4828 J. J. CARRERA-HERNÁNDEZ ET AL.

Figure 8. Contour levels obtained from the total station survey and Structure from Motion. The red
rectangles hightlight zones that show large differences between these methods. images of areas (a),
(b) and (c) are shown on Figure 10.

compare the results obtained with these two methodologies, six different regions are
highlighted on Figure 8 in order to assess the differences between the SfM approach over
the TS survey.
Three of the areas enclosed in Figure 8 are shown in detail in Figure 10: By comparing
Figures 8(a) and 10(a) it can be seen that the SfM survey better represents the terrain;
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4829

Figure 9. Distribution of surveyed points on the total station survey. Coordinates are in UTM-14.

evidently, this was caused by a lack of surveying points on that region, as only a handful of
points were surveyed there Figure 9. As can be seen in Figure 10(a) there is an elevation
difference of approximately 1.5 m between the road and the terrain. The difference
between the contour lines derived from each methodology is more dramatic on
4830 J. J. CARRERA-HERNÁNDEZ ET AL.

Figure 8(f), as the surveyed points were few and only located along one path of that area
(Figure 9); the contour lines derived from the TS survey are linear, which is not the case for
the curves derived from the SfM survey.
Figure 10: images taken on highlighted areas of Figure 8 to show where the UAV-derived
Digital Elevation Model provides better results than a traditional surveying method caused
by undersampling: (a) the terrain is not uniform as is represented from the contour lines
derived from the TS survey – highlighted on rectangle (a) of Figure 8 – (b) steep section not
correctly represented by the total station survey, while (c) and (d) correspond to rectangle
(c) of Figure 8, where the total station survey did not have sufficient surveying points.
Photograph (d) corresponds to the area right above picture (c).
The nearly 3.0 m elevation difference on Figure 8(b) – clearly seen on Figure 10(b) – which
is not reproduced in the TS survey can clearly be seen on the curves extracted from the SfM
survey. In fact, Figure 10(b) shows that this change in elevation is quite steep and is

(a)

(d)

(b) (c)

Figure 10. Images taken on highlighted areas of Figure 8 to show where the UAV-derived Digital
Elevation Model provides better results than a traditional surveying method caused by undersam-
pling: (a) the terrain is not uniform as is represented from the contour lines derived from the TS survey
{ highlighted on rectangle (a) of Figure 8 { (b) steep section not correctly represented by the total
station survey, while (c) and (d) correspond to rectangle (c) of Figure 8, where the total station survey
did not have su_cient surveying points. Photograph (d) corresponds to the area right above picture (c).
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4831

adequately represented in the SfM contour lines. A more dramatic example is perhaps Figure
8(c) which is detailed in Figure 10(c) and (d): the contour lines from the TS survey show a wall
which clearly does not exist (Figure 10(d)); this error was caused by a lack of survey points on
the terrain shown on Figure 10(d), as the points used to develop the contour lines were
measured on the built-up area of Figure 8(c), as can be seen on Figure 9. The contour lines
developed from the SfM survey represent the terrain and even show the wall built around the
building (Figures 8(c) and 10(c)), with an elevation difference of over 3.0 m.
Another difference is shown in Figure 8(d), where the TS survey shows a decline of 1 metre
on the terrain, whereas the SfM survey does not. This decline occurs below the building and is
caused by a small staircase (not shown) which could not be captured by the SfM survey, due
to the fact that – on this area – the terrain is hidden to the UAV camera by the building. The
remaining area used on this discussion is shown in Figure 8(e) – which, by looking at Figure 9
is expected, as no survey points were located on this area, but only on the paved road.
To quantify the difference between both surveys, a Digital Terrain Model (DTM) was
created from the contour lines derived from the TS survey through the r.surf.contour com-
mand of GRASS and subtracted from the DTM generated through the UAV-SfM approach in
order to create a Difference of DEMs (DoDs) as shown in Figure 11, where it can be seen that
errors range from below −2 m to above 2 m. The profile a – a’ of Figure 11 shows how the
contour lines derived from the TS survey misrepresents the outcrop shown on Figure 10(b) by

Figure 11. Differences between Digital Terrain Models derived from the countour lines obtained
through the Total Station survey and through Structure from Motion. Blue colours represent areas
where elevTS > elevSfM, while red colours represent the opposite. This situation is clearly visible on
the profiles a – a’, b – b’ and c – c’.
4832 J. J. CARRERA-HERNÁNDEZ ET AL.

nearly 3 m, and that the contour lines of the TS survey also misrepresent the existing terrain by
as much as 4 m (profile b – b’ of Figure 11), which corresponds to Figure 8(c).
Total Stations provide high accuracy on the sampled points; however, an exhuastive
sampling with TS implicates time and cost – and even an exhaustive survey is prone to
errors caused by not levelling the prism pole or by operator procedures such as centring
and levelling the instrument or pointing at targets (Ghilani 2017). A TS survey is most
advantageously used where high precision of a few (<100) single points is required
(Carrivick, Smith, and Quincey 2016). The large difference observed on Figure 8(c) and
on profile b – b’ of Figure 11) shows that no care was taken to verify the terrain
representation of the contour lines derived from the TS survey, which was probably
caused by the use of an automated interpolation method. The lack of surveying points
on the TS survey on the northwestern region of the domain (Figure 9) hinders the
adequate representation of terrain on that survey – thus producing linear contour lines
on that region.

5. Conclusions
In this work, we report a comparison of two surveying projects which were independently
developed for a heterogeneous built-up area of nearly 33 ha. This comparison was
undertaken to assess if there are any advantages of using the Unmanned Aerial Vehicle
(UAV) and Structure from Motion (SfM) approach over a traditional survey developed with
a Total Station (TS) – in both time and sampling density. The TS surveying crew surveyed
6,240 points over the course of 1 week, while the UAV-SfM survey took approximately 4 h
with a crew of three people – which shows that the UAV-SfM approach is more efficient.
The data generated from both surveys were used to generate contour levels at every
0.5 m throughout the domain, showing that the contour lines derived from the UAV-SfM
approach adequately represent the variation of the elevation on the study area. The spatial
resolution of a TS survey is limited to the spatial distribution of those points surveyed (6,240
for this case), and only provides contour lines which due to the undersampling of the TS
survey do not represent the terrain adequately. The UAV-SfM derived topography is a
subproduct of the resultant three-dimensional point cloud (38  106 points). In addition,
the UAV-SfM topography provides information on its prediction errors (RMSEh = 0.119 m,
RMSEv = 0.129 m in this case).
From the analyses presented, we conclude that the UAV-SfM technology can be
adopted by practising surveyors as the final product is a Digital Elevation Model instead
of a set of points, is less prone to human error and can be undertaken in significantly less
time. In addition, the UAV-SfM approach can be used to survey dangerous and inacces-
sible areas. However, the application of this methodology is limited to flight restrictions of
any Unmanned Air Vehicle.

Acknowledgements
The equipment used in this work was financed by Mexico’s National University (UNAM) through
project PAPIIT-IA108317.
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4833

Disclosure statement
No potential conflict of interest was reported by the authors.

Funding
This work was supported by the Dirección General de Asuntos del Personal Académico, Universidad
Nacional Autónoma de México [PAPIIT-IA108317].

ORCID
J. J. Carrera-Hernández http://orcid.org/0000-0002-4418-9277

References
Agüera-Vega, F., F. Carvajal-Ramírez, and P. Martínez-Carricondo. 2017. “Accuracy of Digital Surface
Models and Orthophotos Derived from Unmanned Aerial Vehicle Photogrammetry.” Journal of
Surveying Engineering 143: 04016025. doi:10.1061/(ASCE)SU.1943-5428.0000206.
Arango, C., and C. A. Morales. 2015. “Comparison between Multicopter UAV and Total Station for
Estimating Stockpile Volumes.” ISPRS - International Archives of the Photogrammetry, Remote
Sensing and Spatial Information Sciences XL-1/W4: 131–135. doi:10.5194/isprsarchives-XL-1-W4-
131-2015.
Artese, S., and M. Perrelli. 2018. “Monitoring A Landslide with High Accuracy by Total Station: A DTM-
based Model to Correct for the Atmospheric Effects.” Geosciences 8: 46. doi:10.3390/
geosciences8020046.
Baiocchi, V., G. Caramanna, D. Costantino, P. J. D’Aranno, F. Giannone, L. Liso, C. Piccaro, A. Sonnessa,
and M. Vecchio. 2018. “First Geomatic Restitution of the Sinkhole Known as ‘Pozzo Del Merro’
(Italy), with the Integration and Comparison of ‘Classic’ and Innovative Geomatic Techniques.”
Environmental Earth Sciences 77: 1–14. doi:10.1007/s12665-018-7244-6.
Buffi, G., P. Manciola, S. Grassi, M. Barberini, and A. Gambi. 2017. “Survey of the Ridracoli Dam: UAV–based
Photogrammetry and Traditional Topographic Techniques in the Inspection of Vertical Structures.”
Geomatics, Natural Hazards and Risk 8: 1562–1579. doi:10.1080/19475705.2017.1362039.
Carrera-Hernández, J., G. Levresse, P. Lacan, and J. J. Aranda-Gómez. 2016. “A Low Cost Technique
for Development of Ultra-high Resolution Topography: Application to A Dry Maar’s Bottom.”
Revista Mexicana De Ciencias Geológicas 33: 122–133. https://www.redalyc.org/pdf/572/
57246037009.pdf.
Carrivick, J. L., M. W. Smith, and D. J. Quincey. 2016. “The Place of Structure from Motion.” In Struct.
From Motion Geosci. chapter 2, 10–36. West Sussex, UK: Wiley & Sons.
Castillo, C., R. Pérez, M. R. James, J. N. Quinton, E. V. Taguas, and J. A. Gómez. 2012. “Comparing the
Accuracy of Several Field Methods for Measuring Gully Erosion.” Soil Science Society of America
Journal 76: 1319. doi:10.2136/sssaj2011.0390.
D’Oleire-Oltmanns, S., I. Marzolff, K. Peter, and J. Ries. 2012. “Unmanned Aerial Vehicle (UAV) for
Monitoring Soil Erosion in Morocco.” Remote Sensing 4: 3390–3416. doi:10.3390/rs4113390.
Darmawan, H., T. R. Walter, K. S. Brotopuspito, Subandriyo, and I. G. M. A. Nandaka. 2018.
“Morphological and Structural Changes at the Merapi Lava Dome Monitored in 2012–15 Using
Unmanned Aerial Vehicles (Uavs).” Journal of Volcanology and Geothermal Research 349: 256–267.
doi:10.1016/j.jvolgeores.2017.11.006.
Dodd, M. 2011. “Where are My Quadrats? Positional Accuracy in Fieldwork.” Methods in Ecology and
Evolution 2: 576–584. doi:10.1111/j.2041-210X.2011.00118.x.
Fonstad, M. A., J. T. Dietrich, B. C. Courville, J. L. Jensen, and P. E. Carbonneau. 2013. “Topographic
Structure from Motion: A New Development in Photogrammetric Measurement.” Earth Surface
Processes and Landforms 38: 421–430. doi:10.1002/esp.3366.
4834 J. J. CARRERA-HERNÁNDEZ ET AL.

Ghilani, C. D. 2017. Elementary Surveying - Introduction to Geomatics. 15th ed. Upper Saddle River, NJ:
Pearson.
Gindraux, S., R. Boesch, and D. Farinotti. 2017. “Accuracy Assessment of Digital Surface Models from
Unmanned Aerial Vehicles’ Imagery on Glaciers.” Remote Sensing 9: 1–15. doi:10.3390/rs9020186.
Girod, L., C. Nuth, A. Kaäb, B. Etzelmüller, and J. Kohler. 2017. “Terrain Changes from Images
Acquired on Opportunistic Flights by SfM Photogrammetry.” Cryosphere 11: 827–840.
doi:10.5194/tc-11-827-2017.
Gonçalves, J., and R. Henriques. 2015. “UAV Photogrammetry for Topographic Monitoring of Coastal
Areas.” ISPRS Journal of Photogrammetry and Remote Sensing 104: 101–111. doi:10.1016/j.
isprsjprs.2015.02.009.
Graham, D. J., and N. G. Midgley. 2000. “Moraine-mound Formation by Englacial Thrusting: The
Younger Dryas Moraines of Cwm Idwal, North Wales.” Geological Society, London, Special
Publications 176: 321–336. doi:10.1144/GSL.SP.2000.176.01.24.
GRASS-Development-Team. 2019. “GRASS GIS Software.” Technical Report. Open Source Geospatial
Foundation. http://grass.osgeo.org.
Immerzeel, W. W., P. D. A. Kraaijenbrink, J. M. Shea, A. B. Shrestha, F. Pellicciotti, M. F. P. Bierkens, and S.
M. D. Jong. 2014. “High-resolution Monitoring of Himalayan Glacier Dynamics Using Unmanned
Aerial Vehicles.” Remote Sensing of Environment 150: 93–103. doi:10.1016/j.rse.2014.04.025.
James, M. R., S. Robson, S. D’Oleire-Oltmanns, and U. Niethammer. 2017. “Optimising UAV
Topographic Surveys Processed with Structure-from-motion: Ground Control Quality, Quantity
and Bundle Adjustment.” Geomorphology 280: 51–66. doi:10.1016/j.geomorph.2016.11.021.
Javernick, L., J. Brasington, and B. Caruso. 2014. “Modeling the Topography of Shallow Braided Rivers
Using Structure-from-Motion Photogrammetry.” Geomorphology 213: 166–182. doi:10.1016/j.
geomorph.2014.01.006.
Johnson, K., E. Nissen, S. Saripalli, J. R. Arrowsmith, P. McGarey, K. Scharer, P. Williams, and K. Blisniuk.
2014. “Rapid Mapping of Ultrafine Fault Zone Topography with Structure from Motion.”
Geosphere 10: 969–986. doi:10.1130/GES01017.1.
Leberl, F., A. Irschara, T. Pock, P. Meixner, M. Gruber, S. Scholz, and A. Wiechert. 2010. “Point Clouds:
Lidar versus 3D Vision.” Photogrammetric Engineering & Remote Sensing 76: 1123–1134. doi:0099-
1112/10/76101123.
Mitášová, H., and L. Mitáš. 1993. “Interpolation by Regularized Spline with Tension: I. Theory and
Implementation.” Mathematical Geology 25: 641–655. doi:10.1007/BF00893171.
Myers, D. T., R. R. Rediske, and J. N. McNair. 2019. “Measuring Streambank Erosion: A Comparison of
Erosion Pins, Total Station, and Terrestrial Laser Scanner.” Water 11: 1846. doi:10.3390/w11091846.
Nelson, A., H. Reuter, and P. Gessler. 2009. “DEM Production Methods and Sources.” In Hengl, T. and
Reuter, H.I. (Eds) Geomorphometry Concepts, Software, Appl. Developments in Soil Science, vol. 33,
chapter 3, 65–85. Amsterdam, Netherlands: Elsevier. doi:10.1016/S0166-2481(08)00003-2.
Niethammer, U., M. R. James, S. Rothmund, J. Travelletti, and M. Joswig. 2012. “UAV-based Remote
Sensing of the Super-Sauze Landslide: Evaluation and Results.” Engineering Geology 128: 2–11.
doi:10.1016/j.enggeo.2011.03.012.
Ouédraogo, M. M., A. Degré, C. Debouche, and J. Lisein. 2014. “The Evaluation of Unmanned Aerial
System-based Photogrammetry and Terrestrial Laser Scanning to Generate DEMs of Agricultural
Watersheds.” Geomorphology 214: 339–355. doi:10.1016/j.geomorph.2014.02.016.
Parsapour-Moghaddam, P., and C. D. Rennie. 2018. “Influence of Meander Confinement on Hydro-
morphodynamics of a Cohesive Meandering Channel.” Water (Switzerland) 10. doi:10.3390/
w10040354.
Ryan, J. C., A. L. Hubbard, J. E. Box, J. Todd, P. Christoffersen, J. R. Carr, T. O. Holt, and N. Snooke. 2015.
“UAV Photogrammetry and Structure from Motion to Assess Calving Dynamics at Store Glacier, a
Large Outlet Draining the Greenland Ice Sheet.” Cryosphere 9: 1–11. doi:10.5194/tc-9-1-2015.
Salvini, R., G. Mastrorocco, M. Seddaiu, D. Rossi, and C. Vanneschi. 2017. “The Use of an Unmanned
Aerial Vehicle for Fracture Mapping within a Marble Quarry (Carrara, Italy): Photogrammetry and
Discrete Fracture Network Modelling.” Geomatics, Natural Hazards and Risk 8: 34–52. doi:10.1080/
19475705.2016.1199053.
 INTERNATIONAL JOURNAL OF REMOTE SENSING 4835

Sona, G., L. Pinto, D. Pagliari, D. Passoni, and R. Gini. 2014. “Experimental Analysis of Different
Software Packages for Orientation and Digital Surface Modelling from UAV Images.” Earth Science
Informatics 7: 97–107. doi:10.1007/s12145-013-0142-2.
Szeliski, R. 2011. “Computer Vision.” In Texts in Computer Science, 812. London, England: Springer.
doi:10.1007/978-1-84882-935-0.
Teza, G., A. Pesci, and A. Ninfo. 2016. “Morphological Analysis for Architectural Applications:
Comparison between Laser Scanning and Structure-from-Motion Photogrammetry.” Journal of
Surveying Engineering 142: 04016004. doi:10.1061/(ASCE)SU.1943-5428.0000172.
Tonkin, T. N., and N. G. Midgley. 2016. “Ground-control Networks for Image Based Surface
Reconstruction: An Investigation of Optimum Survey Designs Using UAV Derived Imagery and
Structure-from-motion Photogrammetry.” Remote Sensing 8: 16–19. doi:10.3390/rs8090786.
Tonkin, T. N., N. G. Midgley, D. J. Graham, and J. C. Labadz. 2014. “The Potential of Small Unmanned
Aircraft Systems and Structure-from-motion for Topographic Surveys: A Test of Emerging
Integrated Approaches at Cwm Idwal, North Wales.” Geomorphology 226: 35–43. doi:10.1016/j.
geomorph.2014.07.021.
Turner, D., A. Lucieer, and L. Wallace. 2014. “Direct Georeferencing of Ultrahigh-resolution UAV
Imagery.” IEEE Transactions on Geoscience and Remote Sensing 52: 2738–2745. doi:10.1109/
TGRS.2013.2265295.
Venter, M. L., C. R. Shields, and M. D. Cuevas Ordóñez. 2018. “Mapping Matacanela: The
Complementary Work of Lidar and Topographical Survey in Southern Veracruz, Mexico.”
Ancient Mesoamerica 29: 81–92. doi:10.1017/S0956536117000128.
Westoby, M. J., J. Brasington, N. F. Glasser, M. J. Hambrey, and J. M. Reynolds. 2012. “‘Structure-from-
motion’ Photogrammetry: A Low-cost, Effective Tool for Geoscience Applications.”
Geomorphology 179: 300–314. doi:10.1016/j.geomorph.2012.08.021.
Wolf, P. R. 2003. “Surveying and Mapping: History, Current Status, and Future Projections.”
Perspectives in Civil Engineering: Commemorating 150th Anniv. Am. Soc. Civ. Eng 128: 163–191.
doi:10.1061/(ASCE)0733-9453(2002)128:3(79).
Woodget, A. S., P. E. Carbonneau, F. Visser, and I. P. Maddock. 2015. “Quantifying Submerged Fluvial
Topography Using Hyperspatial Resolution UAS Imagery and Structure from Motion
Photogrammetry.” Earth Surface Processes and Landforms 40: 47–64. doi:10.1002/esp.3613.
Zhang, K., S. C. Chen, D. Whitman, M. L. Shyu, J. Yan, and C. Zhang. 2003. “A Progressive
Morphological Filter for Removing Nonground Measurements from Airborne LIDAR Data.” IEEE
Transactions on Geoscience and Remote Sensing 41: 872–882. doi:10.1109/TGRS.2003.810682.