PE&RS April 2015 - page 304

system onboard a fixed-wing and a helicopter to analyze vegeta-
tion coverage in bare ground. The data provided by the UAVs
were compared against field estimates, showing good agree-
ment for the measurement of bare ground, particularly with the
helicopter. Breckenridge et al. (2012) estimated the percentage
of coverage in six different types of vegetative (live and dead
shrub, grass, forbs, litter, and bare ground). The final goal in the
above three works was to analyze the ecosystem sustainability.
Suzuki et al. (2010) determined the vegetation coverage
through visible and infrared cameras on-board a fixed-wing
UAV that allows building images mosaics with the help of the
GPS and the IMU. Strecha et al. (2012) proposed the combina-
tion of NIR and visible images to produce orthoimages and
DEMs with machine vision systems installed on UAVs. The goal
was to build vegetation maps to monitor different species.
An unmanned helicopter, with 0.57 kg weight capable
of lifting an 11.5 kg payload with a fuel engine, was used in
Xiang and Tian (2011) to monitor turf grass glyphosate appli-
cation based on multi-spectral CMOS sensor consisting of three
bands (green, 520 nm - 620 nm, red 620 nm - 750 nm, and
NIR 750 nm - 950 nm). The UAV was equipped with additional
sensors: GPS, IMU, video-transmitter, wireless router, single
board computer, and flight controller.
The third topic considered is vegetation development
from different points of view. Imagery-based data from UAVs,
combined with satellite information, were used for studying
vegetation development in braided areas in the French Alps.
The flow of rivers and wind are relevant for the transport of
seeds (Hervouet, 2011).
Lucieer et al. (2010, 2011, 2012, and 2014c) and later
Turner et al. (2014a) in a broad and extensive research, com-
piled in such references, applied SfM and SIFT techniques for
3D mapping of moss beds in Antarctica to determine their
extent along the terrain. They also obtained a thematic map
of moss health derived from the multispectral mosaic using a
Modified Triangular Vegetation Index (MTVI) and an indicative
map of moss surface temperature. For these tasks, they used
two UAVs depending on the goal to be achieved: (a) an electric
helicopter capable of lifting 1.5 kg; and (b) an autopilot octo-
copter with payload of 1 to 1.5 kg. Three sensors which were
used sometimes individually and sometimes all together: (a)
visible digital camera weighing approx. 355 g; (b) multispec-
tral six-band sensor with wavelengths at 530 nm, 550 nm,
580 nm, 670 nm, 700 nm, and 800 nm, determined by 10 nm
filters; and (c) a FLIR thermal sensor. A motion compensated
gimbal mount stabilizes these devices. In addition, the vigor
of these moss beds was addressed in Lucieer et al. (2014b)
using the HyperUAS (see the Multispectral and Hyperspectral
subsection). This was achieved by studying the photosyn-
thetic activity based on the ability to acquire images with
high spatial and spectral resolutions. Different experiments
were conducted to assess the performance in georeferencing
and ortho-rectification that allows the generation of precise
mosaics and DSMs. This HyperUAS allows the quantification of
chlorophyll content and biomass in pasture and barley crop
by computing optical vegetation indices with relevant traits,
namely: NDVI, transformed chlorophyll absorption in reflec-
tance index (TCARI ), and optimized soil-adjusted vegetation
index (OSAVI ).
Negative impact analysis relating to the invasion of grass-
land by woody shrubs was addressed in Rango et al. (2011)
with the aim of reversing this situation to the original one by
analyzing historical data. In this application, UAVs turn out to
be an excellent tool where ranchers and scientists have ex-
ploited their abilities. A fixed-wing UAV was used, equipped
with a consumer visual digital camera in the wing and a
video camera in the nose. The UAV is catapult launched using
a gasoline engine.
Bueren et al. (2015) tested and compared four (visible, vis-
ible + IR, multispectral, and hyperspectral) sensors onboard
two octo-copters to capture the reflectance from grasslands.
Their challenges and limitations were discussed.
Feng et al. (2015) have analyzed urban vegetation using
random forest and texture analysis. Off-the-shelf RGB digital
cameras, onboard a fixed-wing UAV with 2.5 m wingspan and
a length 0f 1.58 m, were used.
Photogrammetry
Photogrammetry is a traditional topic in combination with
remote sensing. UAVs equipped with sufficient sensor tech-
nologies are able to obtain remote measurements from images,
which are conveniently processed in order to produce: 3D
terrain mapping with Digital Elevation Models (DEMs) or DSMs
with shapes, surface reconstruction, elevation contours, or
features. All these products are useful for cartography and
topography, where ortho-images are also final or intermedi-
ate products. Generally, photogrammetry is the support for
other applications. Sometimes, the quality of photogram-
metric products, such as orthophotos, is not always achieved
successfully because of the movement of the UAV or due to
overlapping errors that require special treatments (Samad et
al., 2013). In this regard, studies conducted toward validation
of measurements with robust processing methods are useful
to determine and obtain sufficient quality (Rieke et al. , 2011;
Mesas-Carrascosa et al., 2014, Ai et al. 2015).
Figure 10 displays a DSM built from images captured with
a visible camera (Courtesy of QuantaLab-IAS-CSIC , Cordoba,
Spain). The goal of photogrammetry with respect to UAVs is to
achieve similar or higher accuracies to the ones obtained with
airborne-based systems (Haala et al. , 2011; Remondino et al.,
2011; Strecha, 2011; Liu et al. , 2011) where rapid rotational
or translation movements in UAVs increase the difficulty of
image orientation for subsequent processing, requiring precise
registration or matching techniques. Colomina and Molina
(2014) provide an extensive review related to data processing
techniques considering that, in the context of photogramme-
try from UAVs, their performance is similar to products from
piloted, airborne-based systems. In this regard, the revision
is focused on three main topics: (a) image orientation for
navigation and camera calibration to cope with the problem
of irregularity of frames acquired from UAVs, where computer
vision techniques provide some solutions, such as SfM or
automatic tie point generation based on point detection and
descriptors with sufficient accuracies like SIFT, their variants
and many others; and (b) surface reconstruction, to obtain
DSM and ortho-photos with sufficient accuracies in point
cloud generation and densification, including multi-view
stereopsis techniques.
In the context of photogrammetry, some National Mapping
and Cadastral agencies have considered and acquired UAVs to
develop some activities and products within its competence
(Cramer et al. , 2013). A brief list of some examples is as fol-
lows: spatial data infrastructures, geodesy, GIS, cartography,
topographic mapping, cadastral applications, mapping for
emergencies, erosion, or change detection. In this regard, Eyn-
dt and Volkmann (2013) reported that UAVs suitably equipped
can accomplish these tasks well. Also, Mesas-Carrascosa et al.
(2014) analyzed the potential use of very high resolution UAV
imagery to measure the area of land plots to monitor land pol-
icies. The fixed-wing UAV, with a 2 m wingspan and TOW 5.8
kg was operated by people of the QuantaLab-IAS-CSIC (2014)
team and equipped with a six-band multispectral camera.
Object reconstruction and modeling are also feasible from
remotely sensed data. An important issue addressed in pho-
togrammetry is the validation of measurements from images
for achieving the maximum accuracy possible (Gini et al. ,
2013). Feature detectors (SIFT) and image matching techniques
304 April 2015 PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING