A Comparative Assessment of the
Photogrammetric Accuracy of Mapping Using
UAVs with Smart Devices
Hohyun Jeong, Hoyong Ahn, Dongyoon Shin, Yushin Ahn, and Chuluong Choi
Abstract
This article evaluates
UAV
photogrammetry systems using
smartphones and smart cameras. Image triangulation was
conducted in accordance with interior orientation param-
eters, determined by camera self-calibration. Precise ortho-
mosaic images and digital surface models were generated,
and their accuracy was assessed using aerial and terrestrial
lidar data. Digital surface models were used to estimate
earthwork volumes and verify the suitability of
UAV
pho-
togrammetry for use on construction sites. Georeferencing
accuracy shows that the smart camera performs about twice
as well as the smartphone with reference to checkpoints and
polygon parts. Considering rolling shutter in the smartphone,
it is possible to increase accuracy. Especially in inclined and
rugged topography, the smartphone can benefit from apply-
ing the rolling-shutter method. Earthwork of volume error
is often applied as a legal requirement for some countries,
and our findings indicate that a smart camera with a drone
can be effectively and economically used in earthworks.
Introduction
In their current state, unmanned aerial vehicle (
UAV
) plat-
forms can use smartphones as a sensor and capture geotagged
images (Boddhu
et al.
2013). Smartphones can operate in
3G
and 4G network environments at any time or location, and are
ubiquitous. If the rolling-shutter problem is solved, three-
dimensional (
3D
) information can be easily gathered using
smartphones as the payload in photogra
(Kim
et al.
2013; Aldrovandi
et al.
2015)
facilitate rapid and precise acquisition o
low-cost cameras and good image size—i.e., 12 megapixels
(Raeva, Filipova and Filipov 2016).
When it comes to
3D
modeling, lidar (terrestrial or air-
borne) produces accurate elevation data; but it might not be
economical nor efficient in small- and local-scale mapping
and
3D
modeling studies. A digital terrain model can also be
generated for relatively small areas using global navigation
satellite system (
GNSSs
), particularly when the terrain is
smooth and high-density ground sampling is unnecessary
(Coveney
et al.
2010).
In recent years,
UAV
photogrammetry has advanced
through structure-from-motion (
SfM
) algorithms (Dandois
et
al.
2015), multiview stereo (Seitz
et al.
2006), and computer
vision, so that it has become one of the most powerful tools in
generating digital elevation models. In addition, user-friendly
commercial software, such as PhotoScan, ASP, and Pix4D,
makes
3D
model generation easier than ever. These programs
yield similar
3D
reconstruction results. However, general
assessments state that the
3D
reconstructions of the models
are not always sufficient for photogrammetric applications
(Tscharf
et al.
2015); details including the sides of buildings,
sharp corners, and edges may not be accurately rendered
(Schwind 2016); and not all software packages provide suffi-
cient information for judging a reconstruction’s completeness
(Rumpler
et al.
2017).
A good amount of work has been done to compare differ-
ent software applications and their performances, but smart-
phone applications have not been fully explored.
Smartphone cameras not only cost less than existing pho-
togrammetric
UAV
systems but provide high-resolution images
and
3D
location and attitude data using a variety of built-in
sensors. We used smartphones to conduct an experiment on
the practical applicability of smartphone drones, which have
undergone significant advancement in recent years.
In this study, the focuses are on the smartphone as a pay-
e rolling-shutter effect; comprehensive
sing a
GNSS
receiver, airborne laser
rrestrial laser scanning (
TLS
) data; and
earthworks accuracy and its requirements.
Methodology
Our focuses are on the rolling-shutter effect of the Samsung
Galaxy S7 smartphone, the performance of the Samsung
Galaxy
NX
mirrorless camera, and comprehensive uncertainty
assessment methods using a
GNSS
survey,
TLS
, and
ALS
data .
First, lens-distortion patterns with the S7’s rolling-shutter
option on and off are sought and compared with those of the
mirrorless
NX
camera. After analyzing ground-control-point
(
GCP
) and checkpoint (
CP
) uncertainties in bundle block adjust-
ment, we demonstrate three types of quality verification for
camera-derived orthoimages and digital surface models (
DSMs
):
• Polygon area comparison: Manholes are delineated by
screen digitization on the orthoimage and field survey. The
Hohyun Jeong is with the Korea Land and Geospatial
InfomatiX Corporation, Spatial Information Research
Institute, Iseo-myeon, Jeollabuk-do, Republic of Korea.
Hoyong Ahn is with the Rural Development Administration,
National Institute of Agricultural Sciences, Iseo-myeon,
Jeollabuk-do, Republic of Korea.
Dongyoon Shin is with the National Disaster Management
Research Institute, Ulsan, Republic of Korea.
Yushin Ahn is with the Department of Civil and Geomatics
Engineering, California State University, Fresno, Fresno, CA
93740.
Chuluong Choi is with the Department of Spatial Information
Engineering, Pukyong National University, Busan, Republic of
Korea (
).
Photogrammetric Engineering & Remote Sensing
Vol. 85, No. 12, December 2019, pp. 889–897.
0099-1112/19/889–897
© 2019 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.85.12.889
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
December 2019
889
