The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
1197
Phase I - Remote Sensing vs. Ground Truth (Row 3, 2006)
Li
Remotely S«nsadR«sults (NOVI)
Ground Tr oth (Damaged Leaves (%]) J
Figure 4: Percentage of damaged leaves - Remotely sensed
results (solid green line) and reference data (red dashed line).
3. MSMS - PROJECT PHASE II
After project phase I had shown very encouraging and robust
results, despite the use of an improvised and unfavourable
sensor constellation it was decided to develop a prototype of a
low-cost light-weight airborne multispectral sensor, which
could ideally be flown on the latest generation of rotary or fixed
wing mini or even micro UAVs.
3.1 Sensor Platforms: Micro UAV and Mini UAV
The latest generation of quadcopter micro UAVs with vertical
take-off and landing (VTOL) capability and with maximum
payloads currently in the range of 200g and expected to be in a
range of 1kg could provide ideal remote sensing platforms for
local applications such as agronomical field tests and the
management of specialty crops. In our case, the micro UAV
'microdrones md4-200' (see Figure 5) served as target platform
for the MSMS sensor. The md4-200 is an electrically powered,
GPS/INS-equipped quadcopter with an official maximum
payload of 200g, an unofficial maximum payload of approx.
350g, and a maximum flying time of approx. 20 minutes.
Figure 5: Quadcopter micro UAV 'microdrones md4-200'
with the prototype MSMS multispectral sensor.
3.2 The MSMS Sensor
Based on the earlier results and on the availability of sensor
hardware components at the start of the development, the
design decisions for a low-cost, low-weight MSMS sensor were
as follows: a modular multi-camera concept (see Introduction)
with one camera per band to be sensed - initially limited to the
two bands Red and NIR with the option to extend the number of
channels by incorporating additional camera heads;
panchromatic full frame sensor elements (with the option to
upgrade to higher resolution sensor elements as they become
available); identical, high-grade but low-weight lenses for all
sensor heads; interference filters for the selection of the desired
spectral bands; use of a programmable camera controller with
support for on-board storage of the acquired imagery. The main
features of the current MSMS prototype sensor are:
• two cameras with full frame CMOS sensor elements
(sensor heads MT9V022m integrated into CanCam), 752 *
480 pixels per channel with global shutter (Company:
Feith Sensor to Image)
• CanCam controller with CPU Motorola Coldfire MCF5272
66 MHz and pCLinux (Feith)
• Light-weight C-mount lenses, focal length 8.0mm, F1.3,
interference filters with central wavelengths of 650 nm (R)
and 880 nm (NIR) and a full width-half maximum (FWHM)
of 80 nm and 50 nm respectively
• total weight of the MSMS prototype: 350 g (including
controllers, sensor heads, and the custom-built light-weight
camera frame; sensor powered by UAV battery)
3.3 Field Test Campaign
Due to supply difficulties and an approaching end of the
vegetation season only one test flight campaign could be carried
out so far with the described combination of the MSMS sensor
and the md4-200 platform (17 th of August 2007). The test flight
was again carried out over a grapevine field at the Syngenta test
field in Stein. Since the current sensor exceeds the official
payload limit and since the unofficial payload communicated
by the manufacturer of the UAV turned out to be too optimistic,
data acquisition was only possible in the absence of any wind
and the acquired imagery was limited to a part of the field only.
Figure 6: One of the first MSMS scenes (part) with radiometric
calibration targets (large) and ground control points (small).
3.4 Preliminary Results
Due to the mentioned difficulties in acquiring the first imagery,
there were again a number of challenges in processing the data.
However, these challenges were mainly caused by the very
irregular constellation of the acquired imagery covering only
parts of the area of investigation. Due to the use of a multi
camera payload, the processing chain can principally be
simplified in comparison to the processing steps used in phase I.
Namely the step of a true orthoimage production will no longer
be needed. Early results support the findings from phase I and
again show a strong correlation between plant health status
obtained via remote sensing with the MSMS prototype and the
ground-truth data from the traditional bonification process.