The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
1194
1.1 (Ultra) High-Resolution Remote Sensing in
Agriculture
In the past, the majority of remote sensing applications in
agriculture were either satellite- or ground-based. Over the last
few years we have seen a rapid increase in airborne remote
sensing due to the proliferation of multispectral digital airborne
sensors (Biihler et al., 2007), (Petrie and Walker, 2007).
Table 1 gives an overview of the different remote sensing
platforms with the typical spatial resolution of their
multispectral channels and with their typical fields-of-view
(FOV). This overview illustrates the current resolution gap at
the cm to dm level which could ideally be filled by miniature
UAVs.
Remote Sensing
Platform
Typical Spatial
Resolution (MS)
Typical Field-
of-View (FOV)
Satellite
2-15 m
10-50 km
Aircraft (piloted)
0.2-2 m
2-5 km
Miniature UA V ?
1-20 cm
50-500 m
Ground-based
< 1 cm
< 2 m
Table 1: Typical spatial resolutions and fields-of-view of
different remote sensing platforms - with a spatial resolution
gap between 1 and 10 cm which could be filled by miniature
UAVs.
The trend towards very high-resolution airborne remote sensing
with spatial resolutions in the range of centimetre to decimetre
is driven by agronomical research, management of speciality
crops and investigations of within-field variability in general.
The shift towards precision, or site-specific, crop management,
and the resulting interest in within-field variability, for example,
has been identified as the most significant change in agriculture
over the last ten to fifteen years (Pinter et al., 2003). Further
potential application areas of very high-resolution UAV-based
remote sensing might be the detection and mapping of plant
diseases such as fire blight or the investigation contaminated
sites.
In the following we will primarily focus on the application area
of agronomical research. However, most of the characteristics,
requirements and conclusions also apply to the management of
specialty crops. The term specialty crops includes fruits,
vegetables, tree nuts, dried fruits, and nursery crops (including
floriculture) (USDA, 2004) as well as grapevines, which were
used in the subsequent investigations.
The characteristics and the subsequent remote sensing
requirements of field tests sites can be summarised as follows:
• very small plot sizes down to one square metre resulting in
a ground sampling distance (GSD) in the order of 5-10 cm
in order to ensure statistically reliable results for each test
plot
• regular frequent observations at weekly intervals and at
short notice in order to observe different phenological
developments or other rapidly evolving phenomena
• the manifold of plant species at a test site and the desire to
find a single solution capable for all vegetation types
• relatively simple, robust and rapid processing procedures
with a high level of automation
1.2 Miniature UAVs as Remote Sensing Platforms
Over the last few years we have seen a tremendous
development of UAV technologies at all conceivable sizes,
from business jet sized UAVs right down to artificial 'flying
insects'. There is also an increasing number of projects with the
aim of using UAVs for remote sensing purposes. These UAV
platforms for civilian remote sensing purposes range from large
UAVs (Coronado et al., 2003), (Herwitz et al., 2002) through
mini UAVs (Johnson et al., 2003), (Eisenbeiss, 2004), (Annen
et al., 2007) to micro UAVs presented in this paper (see Table
2).
Due to the rapid development and the ever increasing number
of new UAV concepts and technologies, it has become a
necessity to try and establish a certain classification for UAVs.
The European Association of Unmanned Vehicle Systems
(EUROUVS) has drawn up such classification of UAV systems,
which we will adhere to in this paper. A good overview and
state-of-the-art of UAV systems which is based on the
EUROUVS classification can be found in (Bento, 2008).
Category
Max.
Take Off
Weight
Max.
Flight
Altitude
Endurance
Data
Link
Range
Micro
< 5kg
250m
lh
< 10km
Mini
< 30kg
150-300m
<2h
< 10km
Table 2: Classification mini- and micro UAV systems
Since our UAV-based remote sensing platform is to be
transportable and to be operated locally under minimal legal
restrictions, candidate platforms are limited to the categories of
mini and micro UAVs (see Table 2). Most mini or micro UAV
systems available today integrate a flight control system, which
autonomously stabilises the platform and supports remotely
controlled navigation. Several systems additionally integrate an
autopilot, which permits autonomous flights based on
predefined waypoints - often in combination with
programmable image acquisition. These flight control systems
are typically based on MEMS (Micro-Electro-Mechanical
System) IMU systems, navigation-grade GPS receivers,
barometers, and magnetic compasses. The different sensor
observations are usually integrated to an optimal flight state
using an EKF (Extended Kalman Filter), which is subsequently
used in the flight controller. For mapping applications, it is also
possible to use this flight control data to geo-register the
captured payload sensor data like still images or video streams.
However, as a result of the utilisation of low weight and low
cost flight control sensors, the achievable direct geo-referencing
accuracy is limited to approx. 5-10 metres (Eugster and Nebiker,
2008).
1.3 Low-weight Remote Sensing Payloads
The use of mini or micro UAVs for remote sensing purposes
introduces a number of constraints on the imaging payloads,
namely limitations in terms of weight, power, and space. In
case of micro UAVs there are also very limited possibilities for
payload stabilisation or for the highly accurate direct sensor
georeferencing. Typical weight limitations for imaging
payloads are approx. 20-30% of total weight of the system, e.g.
approximately 300g in case of 1kg micro UAVs and around 5kg
in case of 25-30 kg mini UAVs. While there are an increasing
number of light-weight imaging sensors for the visible spectrum
and for thermal infrared, the situation is completely different in