The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
1195
the case of multispectral sensors. In the case of imaging sensors
for the visible spectrum, there is now a wide range of light
weight consumer cameras with weights of 150g or even less and
with up to 10 MPixels or more. Unfortunately, these sensors
also come with a number of features which have adverse effects
on photogrammetric or remote sensing applications. These
include limited optical quality, mostly zoom lenses, fully
automatic focussing and image stabilisation which make the
task of camera calibration very difficult. There are a number of
operators providing remote sensing services for agriculture
based on RGB imagery only. While this kind of imagery might
provide valuable visual information to farmers, it is certainly
not suitable to analytically assess vegetation properties due to
the lack of information in the NIR band.
Thermal Imaging - There has been a tremendous progress in the
field of miniature thermal imaging sensors over the last few
years resulting in commercially available thermal imaging
sensors with weights in the order of 120g (FLIR Systems, 2005).
This development was influenced by miniature UAV
technologies and was mainly driven by military applications
such as remote reconnaissance (Kostrzewa et al., 2003) and by
applications in the domain of disaster monitoring, namely forest
fires (Rufino and Moccia, 2005), (Esposito et al., 2006).
Multispectral and Hyperspectral Imaging - The development of
sensors for acquiring high-quality, co-registered multi-channel
imagery in the visible and in the Near Infrared (NIR) bands
poses a number of challenges in terms of optics, sensors, sensor
control and calibration. In their recent overview on airborne
digital imaging technologies (Petrie & Walker, 2007) identify
four different concepts for producing multi-channel imagery
with small-format airborne digital cameras. Among them are
single lens solutions with specialised mosaic filters, multiple
arrays, or beam splitters or with solutions based on multiple
cameras. Among the lightest multispectral camera systems
available are the DigiCAM-H39 (IGI, 2007) with a CIR option
and a total weight of approx. 5 kg and Tetracam's recently
released Multichannel Camera MCA4 with approx. 1.8 kg and
ADC2 with approx. 500 g. The latter sensor appears promising
for operations on mini or even micro UAV, however, the
quoted interval of 12 seconds between two individual images
would only permit stationary image acquisition.
Noteworthy investigations by (Rufino and Moccia, 2005) or
(Johnson et al., 2003) create a hyperspectral line sensor by
combining a monochromatic camera with a spectrograph and
attempt to use it on UAVs. However, it remains questionable, if
and how such line sensors can satisfactorily be used on mini or
micro UAVs with their relative low attitude determination
capability and the lack of payload stabilisation.
1.4 The MSMS Project
The Institute of Geomatics Engineering at the University of
Applied Sciences Northwestern Switzerland (FHNW) has been
active in the research of UAV-based applications on the one
hand and remote sensing on the other for several years. The
MSMS (MultiSpectralMicroSensor) project was launched in
early 2006 for researching technologies and applications
pertaining to UAV-based ultra high-resolution remote sensing
in agriculture. The applied research project unites scientists and
professionals from the fields of geomatics, electronics,
agronomical research and UAV technologies. The goal of the
first two project phases presented below, was first to investigate
the feasibility of remote sensing applications based on mini or
even micro UAVs including the suitability and validity of
processing methods and second to develop a prototype low-
weight multispectral sensor as a key component of a flexible
and easy-to-operate low-cost remote sensing system for
extracting plant state variables.
1.5 Field Tests Site
The results of each of the two project phases were evaluated by
flight campaigns at the agronomical R&D plant of Syngenta
Crop Protection AG in Stein (Canton AG, Switzerland). The
test site is located in the northwest part of Switzerland, near the
German border. At the field test site agrochemical substances
and products are applied to various plant species and crop types,
such as potatoes, soya or a number of specialty crops, namely
grapevines. For our pilot studies, a grapevine field was selected,
which consisted of 10 rows which were again subdivided into
test plots of 2.5 meters in length. In total, 240 test plots were
available. At the time of both test flight campaigns, these plots
had been subject to a ground-based 'bonification' by specialists
of Syngenta, which provided an excellent 'real-world' ground
truth for the following remote sensing experiments.
... .
< r ;
m
Figure 1: Grapevine test field consisting of 240 test plots. Also
visible are the large radiometric targets (left) and the smaller
geometric ground control points (bottom).
2. MSMS - PROJECT PHASE I
In the first project phase in 2006 the feasibility and suitability
of mini or micro UAV based remote sensing for agronomical
research applications were to be investigated and demonstrated.
The key findings of the MSMS project Phase I are summarised
below. For a detailed description of the investigations and
results of Phase I please refer to (Brosi, 2006) and (Annen et al.,
2007).
2.1 Sensor Platform: Mini UAV
The test flight campaign of the first project phase was carried
out using model helicopter based mini UAV of weControl AG
(Zurich) (see Figure 2). The UAV has a rotor diameter of 1.8m,
is powered by a combustion engine and can carry an imaging
payload of approx. 1 kg. The mini UAV was equipped with
weControl's flight control system wePilotlOOO, which allowed a
fully automatic waypoint navigation with a verified accuracy of
approx. 3 m. The system also provides a video data link which
allowed a rough online verification of the acquired imagery at
the ground control station.