del aircraft
4125s
72
53
1.7m
/.4 m
)5m
000 s
ce off
ding
otos
19640
miscalibra-
ped and the
ht of 25m
of the mag-
point flight
) and land-
ional flight
sram pack-
ty of Tech-
yma) of the
> x-,y- and
ma) of the
5°, 0.064 °
ntrol point
ly.
was com-
ver, at the
le was just
ting values
were compared to the bundle block results. The given mean val-
ues of the differences could be found in similar magnitude for
other flights over the same test field. The determined position
offset resulted in values of -1.09 m, 0.36 m, and -1.40 m for the x-
, ¥-» and z-coordinate. With the GPS receiver above the projection
center these values do not come from the mounting (lever arm),
but are rather attributed to the realization of the datum. The roll
and nick angle offset was determined to be -0.12° and -1.92°,
respectively. These quantities can be interpreted as mounting cal-
ibration with respect to their size. The yaw angle showed up a
big offset value of -18.24 ^, which was caused by the previously
mentioned error in the calibration of the magnetometer (cf. Table
2).
o mean
T 0.91? 0.129
p 0.94? —1.92?
y 2.34? —18.24?
X | 0.43m —1.09m
Y | 057m 0.36 m
Z | 0.87m —1.40m
Table 2: Standard deviation c and mean of the differences be-
tween direct and indirect georeferencing.
After the substraction of the mean differences the direct and indi-
rect determined values were further studied. The Figures 8, 9 and
10 present individual comparision plots between the direct and in-
direct georeferencing results. Fig. 8 shows the deviations for the
X coordinate, after subtraction of the mean. The pattern is re-
peated very well, but some larger systematic offsets, e.g. at 350 s,
occur. Fig. 9 shows that the quadrocopter is keeping the height
during the flight more stable, than it is reported by the sensors.
This differences may manly be introduced by air pressure varia-
tions introduced by varying motor speed which was necessary to
stabilize and navigate the UAV at the predefined flight path. Fi-
nally, Fig. 10 shows that the roll angle variation during the flight
is more than 14^ reported by the on board sensors, whereas the
aero-triangulation provides a range of 8 °. The on board sensors,
however, model a large portion of the variability of the reference
data.
The standard deviation for roll and pitch differences, i.e. after
subtracting the mean values from the differences is 0.9 °, and
therefore well within the specifications of the sensors. The yaw
of the camera could be determined less accurate (only integer val-
ues are provided by the magnetometer and, as mentioned before,
the gyroscope observations were not considered during the pro-
cessing of the yaw angle of this data set), but still has an accuracy
of 2°. For the determination of the camera position the stan-
dard deviation is 0.43 m, 0.57 m, and 0.87 m for the x-, y-, and
z-coordinates and is well below the expected accuracy of 1 m (cf.
Table 2). Due to the subtraction of the mean, this estimation may
be too optimistic. However, the evaluation is based on more than
50 photos.
5 CONCLUSIONS
UAV systems are a promising platform for close range airborne
photogrammetry. Within this paper it could be demonstrated that
the accuracy of on board devices of light weight (smaller than
lkg) UAV platforms can be used for direct georeferencing of the
acquired imagery with a positional accuracy (1 sigma) below Im
meter, a nick and roll angle accuracy of smaller than 1 ^ and a yaw
accuracy smaller than 2.5 ^. With the camera's parameters and a
flying height of 25 m above ground (GSD 7 mm), this propagates
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
[m]
19630
19625 L.— F4
d 1 $ ory PS / id 9. zi N
19620 t ; rt. Arr $ d Pee Sy AT
| \ / M e
T
19615 : ' 7 : i ud
19610 - 2 |
19605 - \ ; J
19600 - (ug J
19595 - Vi J
50 100 150 200 250 300 350 400
flight time [s]
Figure 8: Comparison of direct and indirect georeferencing for
the x-coordinate. The red line represents the direct georeferenc-
ing result, whereas the blue dots correspond to the indirect geo-
referencing.
[m]
185
T T T T
/*
ee LUTZ. £5 La
A jue,
^
des A / Va a d^
600% e, ans
180 | / : nd
175 /- d + . à : ; t 4
170
165 + / Ld
160 d. , : T En
i
50 100 150 200 250 300 350 400
flight time [s]
Figure 9: Comparison of direct and indirect georeferencing for
the z-coordinate. The red line represents the direct georeferenc-
ing result, whereas the blue dots correspond to the indirect geo-
referencing.
0 50 100 150 200. 250.300 + = 350. 400
flight time [s]
Figure 10: Comparison of direct and indirect georeferencing for
the roll angle. The red line represents the direct georeferencing
result, whereas the blue dots correspond to the indirect georefer-
encing.
