PE&RS April 2015 - page 305

(least-squares) were evaluated in Lingua et al. (2009) for DSM
generation. The images were acquired with an off-the-shelf
visible light camera.
Progress related to machine learning-based techniques for
DSMs generation has reinforced its use for images acquired
from UAVs (Rosnell et al. 2011).
Virtual reality is another issue closely related to pho-
togrammetry. In this way, Linkugel and Schilling (2013)
proposed a simulation system where a micro-UAV is used
for computing 3D measures for virtual reality purposes. The
mathematical model was described, including all aerodynam-
ic parameters of the UAV toward the definition of the geomet-
ric modeling based on different sensors.
Digital elevation models and 3D mapping with DSM or DTM
production together with mosaicking with geo- and ortho-
rectification are two main topics inside photogrammetry;
both are considered separated here, although they are closely
related.
3D Mapping, Digital Surface, Elevation, and Terrain Models
Nex and Remondino (2014) and Remondino et al. (2011)
provided a review with new insights and proposal for differ-
ent photogrammetry-based applications, including 3D digital
terrain or 3D textured models. Photogrammetric approaches, in-
cluding topographic maps with slopes, have been described in
Tahar et al. (2011 and 2012) oriented to landslides applications.
Hugenholtz et al. (2013) evaluated the accuracy in DTM
production using a fixed-wing UAV that weighs less than 6.2 kg
and equipped with an off-the-shelf CCD-based visible camera.
Different works have been proposed for 3 D model genera-
tion. In this regard, point cloud generation is a task of interest
for 3D mapping accuracy, a procedure for such a purpose was
proposed in Rosnell and Honkavaara (2012) with two RGB-
based digital still CCD-cameras. Two quad-copters were used,
which were able to carry 300 g and 1.2 kg payload, equipped
with cameras weighing 180 g and 448 g, respectively.
Neitzel et al. (2011) used an octo-copter (1.2 kg net weight
or TOW 2 kg with camera) for 3D mapping landfills with the
aim of determining its volume and quantity based on point
cloud computation. 3D building models are obtained in Jizhou
et al. (2004) using a fixed-wing UAV equipped with a CCD-based
camera. They captured oblique images to obtain relevant parts
of buildings instead of using a pair of images as usual.
Harwin and Lucieer (2012a and 2012b) applied multi-
view stereovision (MVS) techniques to obtain 3D structure
from overlapping imagery captured from multiple angles. An
octo-copter with approximate payload limit of 1 kg is the UAV
used. It was equipped with a stabilized camera mount to carry
different sensors, including a commercial digital camera. A
very dense point cloud was produced with sufficient accura-
cy. Accuracy is a central issue in photogrammetry as reported
in Küng et al. (2011a) and Vallet et al. (2011), where different
experiments have been carried out with light UAVs, weighing
less than 500 g with maximum payload of 125 g. Different
methods and strategies for point cloud generation from digital
images captured with UAVs flying at relatively low altitudes
were also addressed in Siebert and Teizer (2014) for 3D map-
ping in mapping earthwork projects.
Regarding digital surface models, a laser scanner, two CCD-
based digital cameras (with weights of 500 g each) and two
infrared devices (NIR sensitivity with 500 g) were integrated
together with an IMU and GPS in Nagai et al. (2009) for such a
purpose. A 3D shape is obtained by the laser scanner as point
cloud data, texture information is acquired by the digital cam-
eras, and vegetation indexes are acquired by the IR cameras
simultaneously. The UAV is a helicopter with weight of 330 kg
and payload of 100 kg with two main rotors (4.8 m diameter)
and two tail rotors (diameter of 0.8 m).
High-resolution surface models are possible by using UAVs
flying at low altitudes. Mancini et al. (2013) developed a
method based on SfM to build such models in unstructured
coastal environments. An electric hexacopter was the UAV
used, with a 1 m diameter and total weight of approximately
5 kg, and equipped with a digital camera. Walker (2012) also
addressed the topic of coastal management applications.
In Delacourt et al. (2009) DEMs and orthorectified images,
acquired from and helicopter, are built with high spatial
resolutions (<5 cm). The system was tested on the beach of
Porsmillin (French Brittany) for the quantification of morpho-
sedimentary changes of the coastal fringe, including cross-
shore and long-shore sediment transport.
Imagery change detection techniques for topographical
reconstruction were applied in Xuan (2011), where DSMs
were built from the remotely sensed data based on different
techniques, including: triangulation, DEM generation, ortho-
imaging, and mosaicking. Two types of UAVs were used: (a)
Figure 10. Digital Surface Model built from images captured with a visible light camera (Image courtesy of QuantaLab-ias-csic, Cordoba, Spain).
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING April 2015 305