Desikan
et al.
(2013) used several
UAVs
to avoid un-authorized
entry in aquatic ecosystem (park) to enhance the conservation
of endangered aquatic species in its natural habitat.
Specifically, target detection and tracking are two impor-
tant challenges in surveillance: often with partial information,
because the target could be only partially observed due to large
areas not covered by the onboard sensors on. This challenge
occurs when tracking is conducted with sensors having a
limited field of view, such as camera-based or lasers. In image-
based applications, the key issue in target tracking is the object
matching between successive frames, so that an object can be
located from one frame to the next. This analysis is not with-
out significant problems in outdoor environments, such as the
ones reported in Kwon
et al.
(2013), where sunlight reflections
on water surfaces become confused with objects. Relevant fea-
tures of objects must be extracted to be matched; in this regard,
super-resolution techniques have been applied in video sur-
veillance for improving images quality (Camargo
et al.
2010).
Al-Helal and Sprinkle (2010) and Siam and ElHelw (2012)
proposed a solution for tracking a target moving in the ground
based on visual detection, where significant features, such as
corners or specific lines are used for guiding the process.
Hong
et al.
(2008) applied a continuous wavelet-based ap-
proach for target tracking in video sequences for surveillance
purposes. They converted target trajectories in a spatial-
temporal domain into target energy volumes in the frequency
domain where different motion parameters were integrated to
obtain three target energy densities, which then serve as cost
functions for estimating target trajectories and sizes.
Vehicle detection, traffic tracking, and monitoring are top-
ics of special interest in surveillance, where
UAVs
can play
an important role. Coifman
et al.
(2006) proposed a roadway
monitoring approach form
UAVs
. Kanistrasy
et al.
(2013)
provided a survey on this area with additional references to
related topics, including road and moving vehicle detection,
vehicle counting, traffic flow and behavior. Liu
et al.
(2012)
used
UAVs
for moving vehicle detection and tracking based on
similarity measurements between consecutive frames in the
video stream. Skoglar
et al.
(2012) proposed a method to track
several vehicles on roads based on vision sensor with gimbal
capabilities. All targets are monitored with simultaneous ac-
tive search, and also for identifying new targets. The vision
sensor is oriented towards the field of interest.
Xiao
et al.
(2008) and Miller
et al.
(2008) proposed sev-
eral techniques for tracking persons in video sequences from
UAVs
. The first work used video cameras for tracking ground
vehicles and the second one used infrared images.
A target tracking approach, based on a monocular camera
(pinhole model), for determining ranges from
UAVs
to objects
was the approach proposed in Choi and Kim (2014). A guid-
ance law is also proposed for such purpose.
Different transformations and approaches have been
proposed for target detection and tracking.
SIFT
and variants
such as the Mean
SIFT
, were proposed with the aim of match-
ing objects in successive frames (Fang
et al.
, 2011; Chao-Jian
and San-Xue, 2011). Gleason
et al.
(2011) described a method
based on Mean
SIFT
for vehicle detection and tracking in rural
areas with the aim of detecting potential threats in oil and gas
pipelines with extension to other applications in these kinds
of environments. Automatic aerial surveillance systems in
buried gas and oil pipelines based on
UAVs
were considered in
Zaréa
et al.
(2014).
Fang
et al.
(2011) proposed particle filtering based on the
mean-shift algorithm captured with an
UAV
. Rodríguez-Canosa
et al.
(2012) described a compensated optical flow-based
approach between consecutive frames for tracking objects,
where low frequency vibrations caused by the
UAV
are con-
veniently balanced. Lin and Saripalli (2012) applied a Hough
transform-based approach for road detection and tracking in
desert environments from images acquired with a commercial
camera onboard an autonomous helicopter. Ruangwiset (2009)
introduced a path planning strategy for target tracking from
UAVs
. A chaotic biogeography-based optimization approach to
target detection has been proposed in Zhang and Duan (2014)
where the chaotic strategy, the dynamic ergodicity popula-
tion and global searching avoid local optimal solutions during
evolution. An inter-row tree tracking technique was applied
in Thamrin
et al.
(2012) based on Structure from Motion (
SfM
).
Environmental Monitoring
Environmental monitoring operations are an interesting area
inside remote sensing, where
UAV
technology has much to
contribute. Research in this field is one of the key pillars
(Hardin and Hardin, 2010). Anderson and Gaston (2013)
provided a broad revision of
UAV
-based applications covering
several areas, where ecology is one of the most relevant. A
review was conducted under the United Nations Environment
Programme (
UNEP
, 2013) where drones are used to work in a
broad variety of ecosystems.
The topics addressed under the environmental monitoring
topic include: volcanic inspections, soil erosion, landslides or
rocky surfaces, aquatic environments, canopy in the understo-
rey, rural roads, and geological hazard analysis.
Volcanic Inspections
Volcanic inspections and monitoring are risky activities
where
UAVs
are appropriate tools (Smith
et al.
, 2009). The
experiments carried out in volcanic areas are intended for
predicting eruptions and for issuing possible warning to resi-
dents. These explorations are on-going, (McGarry, 2005).
NASA
(2015c) flew modified Dragon-eye drones, fixed-wing,
with a weight of 2.7 kg and payload of 500 g, for monitoring
the Turrialba Volcano in Costa Rica with the aim to study gas
emissions and ash clouds inside the volcanic plume (Wil-
liams, 2014). Pieri
et al.
(2014) described how to acquire
different measurements, including gases (CO
2
, CH
4
, H
2
S, He)
and aerosols liquid (H
2
SO
4
, HCl) at the same time atmospheric
data (temperature, humidity, pressure, wind velocity) are
obtained with
UAVs
in the Turrilba Volcano, Costa Rica. In
this line are focused the works described in Mondragón
et al.
(2015) for volcano inspections and hydrothermal alterations
at mountains of Poas and Irazu (Costa Rica) with a multi-rotor
UAV
equipped with thermal and visible cameras. McGonigle
et al.
(2008) conducted different experiments to measure
volcanic gases at La Fossa crater, Vulcano Island, Italy with a
multi-gas sensor (see the Chemical Sensors subsection). This
device was used in Shinohara (2013) for analyzing gas emis-
sions in Shinmoedake, Kirishima Volcano, Japan.
Amici
et al.
(2013) described the inspection of the Le Sa-
linelle, Italian mud volcano on the lower South West flank of
the Etna Volcano. The
UAV
is configured as a hexa-copter with
a 1.7 kg payload, equipped with a lightweight thermal system
of 67 g and spectral response in the range at 2 µm to 14 µm
and a PAL video camera with weight of 600 g.
Soils
Soil erosion analysis, hazard monitoring, reflectance proper-
ties, and 3
D
modeling are motivating applications where
UAVs
can play an important role.
Frankenberger
et al.
(2008) evaluated the feasibility of us-
ing low-altitude (100 m) photogrammetry to assess ephemeral
gully erosion at agricultural fields after rainfall events.
DEMs
were built for analysis from surface images acquired with
a commercial camera onboard an
UAV
and checked against
ground-based systems, such as terrestrial lidar.
Hazard monitoring, for rockslides in Randa (Wallis, Swiss
Alps), was addressed in Eisenbeiss (2009). The images were
acquired with a still video camera on-board an unmanned
helicopter.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
April 2015
301
