CES
. Goad*
U.S.A.
jal targets. The first
ie second application
ors. In both cases the
€ structural dynamic
duced vibration. A
comparison between
targets were tracked
nd offline processing
the photogrammetric
visualisations of the
iper will describe the
> geometry and image
| to track the target
IGHT VEHICLES
etric monitoring of
f the membrane and
icle. These 250 mm
ges of research and
role in military
vell as the subject of
s and materials. The
pment is primarily in
Defense Advanced
develop micro aerial
inches and a speed of
's of wings are under
tions have generally
insparent monofilm
hite/epoxy spars or
is that it will be an
a sensor and radio
) be a vision system
red bands, however
diation counters or
ly. The vehicles will
return intelligence
ved advantages of the
detected, relatively
capable of providing
accurate and valuable information from redundant sources.
Micro-flight vehicles also have civilian applications to search
and rescue or fire fighting operations, for example.
Previous research on micro-flight vehicles has concentrated on
fixed wing designs with testing carried out in wind tunnels
(Waszak et al, 2001). Both qualitative (flow visualisation) and
quantitative (target tracking from a single, high frame rate
camera) videometric techniques have been used to characterise
the deformation of the membrane surfaces under wind load and
at different angles of attack (Waszak et al, 2001). The new
concept under development at Langley is a flapping wing type
vehicle with a self-contained power source, avoiding the noise
or exhaust trail associated with conventional propulsion
systems. The flapping of the wings is generated by a simple,
low speed electric motor with an off-centre counterweight to
induce flapping indirectly.
The measurement task for the micro-flight vehicle prototype
was to characterise the motion of the wing surfaces at different
flapping frequencies and to compare two membranes of
differing thickness. The set-up to capture imagery is shown in
figure 1. The wings and flap motor are mounted vertically on
an optical table, opposite two Hitachi monochrome CCD
cameras, also mounted vertically. The configuration was later
changed to a horizontal mounting of the cameras to improve the
sensitivity of the tracking. The Hitachi cameras generate RS-
170 analog video output with a resolution of 752 by 480 pixels,
captured in this case by dual Epix frame grabber cards. The
cameras were locked together using a master-slave link through
external synchronisation from one camera to the other. Ring
lights were used to illuminate the retro-targets, placed both on
the fixture, to provide a fixed reference, and on the spars of the
wings, to determine the shape of the wings. Retro-targets were
not placed on the membrane sections as the material is very
delicate and because the spars control the overall shape of the
surface.
The camera set-up was calibrated using the small 3D target
array seen lying on the optical table in figure 1. This fixture
was moved around within the field of view of the cameras to
simulate a multi-exposure, convergent photogrammetric
network. The network for the simultaneous calibration of the
two cameras comprised 44 targets and 40 full frame exposures.
The relative precision of the network was 1:32,000,
corresponding to a mean coordinate precision of several
micrometres for the targets. The camera calibration and the
relative orientation of the two cameras were derived from the
network. This technique uses post-processing of the camera
station data to determine the base vectors and relative rotations
of the two or more cameras, and has been used successfully for
underwater stereo-video (Harvey and Shortis, 1996) and wind
tunnel testing (Shortis and Snow, 1997) applications.
Fields, rather than frames, were captured during the
measurement process to increase the sample rate to 60Hz. The
calibration data sets for the cameras were converted from
frames, captured during the calibration process, to fields,
captured during the measurement process, using the known
relationship between frames and odd/even fields (Shortis and
Snow, 1995). The use of fields, rather than frames, halves the
vertical resolution of the exposures. However the aspect ratio of
the object allowed the CCD sensors to be aligned with the
horizontal axes parallel to the general direction of motion
-9] —
Figure 1. Initial experimental set-up for the calibration and
measurement of the wing surfaces of the micro-flight vehicle.
within the images, thereby minimising any loss in system
sensitivity.
Left and right images were captured as a sequence of individual
images in TIFF format, and correlated using VITC time code
generator input, as shown on the two images in figure 2. A
number of sequences were captured at different wing flap
Figure 2. Left (top) and right images of the wing surfaces
of the micro-flight vehicle.