Geomorphic Change Detection Using Cost-
Effective Structure-from-Motion Photogrammetry:
Evaluation of Direct Georeferencing from
Consumer-Grade UAS at Orewa Beach
(New Zealand)
Stephane Bertin, Benjamin Levy, Trevor Gee, and Patrice Delmas
Abstract
Unmanned aerial systems (
UAS
) and structure-from-motion
photogrammetry are transforming the way we produce topo-
graphic data, with applications covering many disciplines
in the geosciences, including coastal studies. To overcome
limitations of ground control points (
GCPs
), we evaluate
direct georeferencing (
DG
) of consumer
UAS
imagery for the
cost-effective measurement of beach topography. Using
DG
,
camera positions determined with on-board instruments
provide air control points for photogrammetry, obviating
the need for presurveyed
GCPs
. We validate the approach at
Orewa Beach, New Zealand, achieving vertical accuracies
similar to light detection and ranging (< 0.2 m) at a higher
resolution (< 0.1 m). A low-quality global navigation satellite
system onboard a consumer
UAS
remains the main constraint
on measurement quality. We show how independent topo-
graphic data sets, which are increasingly available world-
wide, can improve measurement quality, and hence change
detection capacity. Our understanding of measurement
quality achieved in this study is applied to the assessment
of morphological and volumetric change at Orewa Beach.
Introduction
The coast is a dynamic and fragile envir
ments of the coastal zone should ideally reflect this temporal-
ity if we aim to understand the (geomorphic) processes and
changes that occur in response to natural and anthropogenic
drivers. Developments in topography remote sensing (e.g.,
image-based and laser scanning techniques) have greatly im-
proved our ability to survey the coast at a large scale allowing,
for example, to transition from labor-intensive beach profiles
to digital surface models (
DSMs
), with higher resolution and
surface coverage. In New Zealand, airborne lidar (light detec-
tion and ranging) was used to survey the Auckland isthmus
in 2013, resulting in topographic data covering over 2000
km
2
and including most of the coastal zone (
LINZ
2019). Such
data sets are increasingly valuable for detailed coastal flood
risk analyses (Raji
et al.
2011), providing coastal managers
with essential information for adapting coastal areas to storm
events and rising sea levels (e.g., Ministry for the Environ-
ment 2015; National Institute of Water and Atmosphere 2015).
However, the cost for lidar data acquisition remains high,
and thus seldom enables repeated measurements to monitor
coastal change, such as rapid storm-induced erosion or even
annual/pluriannual evolution.
A growing alternative in topographic remote sensing is
the use of (small) unmanned aerial systems (
UAS
), commonly
referred to as drones, equipped with photographic sensors.
With cost-effective technologies that continue to improve,
UAS
are increasingly accessible and versatile, allowing the capture
of aerial imagery in a variety of situations. Compared with
full-scale airborne platforms,
UAS
provide the opportunity of
frequent data capture at a very low cost, with sampling peri-
ods that can be virtually as low as minutes or hours. Further-
more, adapting flight plans and image acquisition geometries
to the terrain and study needs has become relatively straight-
forward using drones, which can improve survey resolution
and precision. Parallel developments in Computer Vision, for
previous constraints on stereo corre-
ucing uncertainty in photogrammetry
well 2000) have resulted in the rapid
hotogrammetry in scientific studies.
Particularly, structure-from-motion (
SfM
) and multi-view ste-
reo algorithms (Lowe 2004; Furukawa and Ponce 2010) allow
the production of high-quality
DSMs
and orthophotographs
from imagery collected with different orientations and scales
using consumer (nonmetric) cameras. This photogrammetric
approach is commonly referred to as
SfM
photogrammetry.
Although using
UAS
and
SfM
photogrammetry in combina-
tion is becoming widespread in scientific studies to measure
topography (see, e.g., the review by Colomina and Molina
2014), with several applications to the coastal zone (e.g.,
Mancini
et al.
2013; Ierodiaconou, Schimel, and Kennedy
2016; Long
et al.
2016; Turner, Harley, and Drummond 2016;
Talavera
et al.
2018; Laporte-Fauret
et al.
2019), we argue that
further methodological development is necessary to the larger
uptake of the technique and increased accessibility, including
outside academic disciplines. In particular, methodologies
Stephane Bertin is with the Institut Universitaire Européen de
la Mer (IUEM), Univ. Brest, Laboratoire Géosciences Océan-
UMR 6538, Technopôle Brest-Iroise, Rue Dumont d’Urville,
F-29280 Plouzané, France (
).
Stephane Bertin and Benjamin Levy are with EPhoS Limited,
Lorne Street, NZ-1010 Auckland, New Zealand.
Trevor Gee and Patrice Delmas are with the Department of
Computer Science, The University of Auckland, Princes
Street, NZ-1010 Auckland, New Zealand.
Photogrammetric Engineering & Remote Sensing
Vol. 86, No. 5, May 2020, pp. 289–298.
0099-1112/20/289–298
© 2020 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.86.5.289
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2020
289
