commercial digital camera. Leaf area index measurements
were also obtained with successful performance results, com-
parable to the ones obtained by lidar technologies.
Turner
et al.
(2011) provided a method for mapping vine-
yards based on visible, multispectral and thermal imagery.
Efficient water use regulating the irrigation schedule and
determining results of special interest for efficiency concerning
plant bio-physical parameters and physiological status are stud-
ied in forestry. In this regard, Baluja
et al.
(2012) analyzed the
water status variability in vineyards based on multispectral im-
ages (six-bands from 580 nm - 800 nm), including visible and
NIR
bands. Gonzalez-Dugo
et al.
(2013) computed a crop water stress
index (
CWSI
) for determining the water status in an orchard with
an
UAV
, 2 m wingspan, fixed-wing and 5.8 kg
TOW
equipped
with a thermal camera. Gago
et al.
(2014) computed leafs water
stress in an experimental vineyard with an
UAV
hexa-copter
equipped with a thermal camera and an additional
RGB
camera.
Zarco-Tejada
et al.
(2013
b
) computed the
CWSI
in vine-
yards, with the aim of determining different irrigation levels,
using both a multispectral and a thermal camera on board a
2 m wingspan, fixed-wing platform, with a 5.8 kg
TOW
. The
spectral band-set center comprises wavelengths between 530
nm and 800 nm.
CWSI
was also computed in Bellvert
et al.
(2014) based on canopy temperatures measured with infrared
temperature-based sensors placed on top of grapevines to
map the spatial variability in water deficits in a “Pinot Noir”
vineyard.
CWSI
was correlated with leaf water potential in Zar-
co-Tejada
et al.
(2012). This correlation was also tested using
thermal imagery captured with a sensor with spectral range of
8 µm - 12 µm onboard an
UAV
.
Primicerio
et al.
(2012) used a hexa-copter for site-specific
vineyard management based on canopy analysis from
NDVI
and equipped with a multi-spectral
CMOS
-based camera, with
weight of 200 g and wavelengths of 520 nm to 600 nm (green),
630 nm to 690 nm (red), and 760 µm to 900 nm (
NIR
).
A helicopter carrying an imaging payload of approximately
1 kg was used in Nebiker
et al.
(2008) for deriving plant
health, based on the computation of the percentage of dam-
aged leaves within a grape vine.
NDVI
values are obtained by
RGB
(
CMOS
-based) and
NIR
(
CCD
-based) sensors. The
RGB
and
NIR
images are captured on different flights for subsequent
geo-referencing and ortho-rectification.
Gonzalez-Dugo
et al.
(2015) conducted different studies in
a commercial pistachio orchard located in Madera County,
California to determine the spatial variability in water status
and irrigation needs based on thermal imagery computing the
CWSI. The
UAV
platform used is described in Zarco-Tejada
et
al.
(2012 and 2013b).
Plate 3 displays a plot, corresponding to a forestry plantation,
obtained with an
ALTM
Gemini laser scanner, described above.
Forests
In forest or forested areas,
UAVs
are also useful. They allow
flying over the forest stand for different purposes. Canopy
analysis, including gap patterns, and 3
D
mapping are two
relevant areas where
UAVs
are used.
Dunford
et al.
(2009) proposed a classification approach,
based on imaging analysis, to quantify riparian areas and
vegetation in the Mediterranean region, where standing dead
trees are identified as well as unhealthy or dead canopy. Some
potential and limitations were also reported, such as typical
problems derived from sizes and payloads of
UAVs
or illu-
mination or vibrations/undesired movements in the sensor
during flight. Advanced technologies have been designed for
minimizing such effects, including cameras with auto-iris,
gimbals, or three-axis stabilized platforms.
Whalin (2012) proposed the analysis of tree canopies to de-
tect the health based on infrared analysis. This allows detect-
ing and fighting beetle pests as well as other diseases affecting
forests, which can also be monitored.
Wallace
et al.
(2012a) used a lidar scanner, together with
an
IMU
,
GPS
, and high-resolution visible video cameras
onboard an octo-copter for tree height estimation and forest
inventory. A very high-density point cloud (up to 62 points
per m
2
) is achieved for the measurement of tree location, as
well as height and crown width, which were assessed over
individual isolated trees. Lidar was also the technology used
in Wallace
et al.
(2011, 2012b, 2014a, and 2014b) for forest
inventories based on change-detection analysis with high per-
formance, including error assessment. The stability of canopy
maps in forested areas was analyzed in Wallace (2013) based
on a lidar system onboard an
UAV
. Wallace
et al.
(2014b and
2014c) used laser-based technology onboard an octo-copter
in a four-year-old Eucalyptus globulus stand for determining
stage of growth and the rate of pruning, respectively, with the
aim of achieving high quality timber. The laser consists of
four parallel scanning layers each with a scan frequency of 12
Hz, being capable of recording up to three returns per pulse
with a transversal beam divergence of 0.8°.
Hernández-Clemente
et al.
(2012) used multi (hyper) spec-
tral imagery to obtain biochemical (chlorophylls, carotenoid,
xanthophyll) measurements in forest canopies with conifers.
The
UAV
was a 2 m fixed-wing platform capable of carrying a
3.5 kg payload. The camera consisted of six independent im-
age sensors and optics with user configurable spectral filters.
Tree height canopy measurements were obtained in Zarco-
Tejada
et al.
(2014) from a
RGB
camera manipulated to capture
the near-infrared spectral band, i.e., a
CIR
device previously
introduced.
Dandois and Ellis (2010 and 2013) obtained high-resolu-
tion 3
D
maps in forestry vegetation from
RGB
images captured
with an
UAV
. They achieved similar performance as was
obtained with lidar systems. In this regard, Tao
et al.
(2010)
computed dense point clouds from images captured with
UAVs
for 3
D
mapping purposes.
A mini-
UAV
-borne lidar system was constructed in Lin
et
al.
(2011). The
UAV
is a helicopter with weight of 4.5 kg, being
able to transport a payload of about 7 kg. It is equipped with
laser scanners of 1.2 and 1.6 kg for assessing its validity in
high-resolution 3
D
mapping for tree height estimation.
Fritz
et al.
(2013) used an octo-copter for tree stem detec-
tion in open stands. The
UAV
is equipped with a consumer
camera fixed on a flexible mount, which enables tilting the
camera vertically and horizontally. In five steps, a dense point
cloud is generated: Scale Invariant Feature Transform (
SIFT
)
operator for generate tie points, image matching of
SIFT
fea-
tures, bundle adjustment to estimate camera parameters, clus-
tering the image, and dense reconstruction. The method was
validated against point clouds from terrestrial laser scanners.
Chisholm
et al.
(2013) used a quad-copter equipped with a
lidar operating at 10 Hz with 1,081 beams per scan, with scan-
ning angle of 270° and range of 30 m for below-canopy surveys.
A map of horizontal cross section of the forest was reconstruct-
ed and the diameter-at-breast-height of 12 trunks estimated.
Fallen trees are surveyed from an unmanned helicopter in
Inoue
et al.
(2014) in a deciduous broadleaved forest in east-
ern Japan as a key factor in biodiversity and biogeochemical
cycling. The
UAV
was equipped with a consumer-grade digital
camera a
GPS
and a laser range finder for the production of
DEMs
.
Other Applications
Soil monitoring in agriculture becomes an important task be-
cause yield can be estimated based on its evaluation, wherein
UAVs
can play an important role. Biasio
et al.
(2010) used
multispectral imagery (a device with three visible and two
infrared channels) to monitor the soil composition in agricul-
tural fields to estimate crop yields based on the computation
of vegetation indices in farmlands.
Corbane
et al.
(2012) addressed the study of soil surface
characteristics in vineyards with the aim of determining
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
April 2015
297
