The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008
1210
“SUSI” took place on 23.05.2007. Changing winds prohibited a
systematic aerial survey, thus several attempts were necessary
to cover the area appropriately.
4.2 Photogrammetric workflow
The workflow of the photogrammetric data processing from
single images to an orthophoto mosaic by using the software
Leica Photogrammetry Suite (LPS) 9.1 is shown in
Figure 4.
The primary goal was a more or less automated workflow. Due
to several problems during the aerial surveys, it turned out that
several manual and semiautomatic steps are required in the
processing chain of the data,
Figure 4. For instance the autonomous navigation did not work
properly which resulted in rather unsystematic strips. In turn
after the flight the best image data from different strips had to
be selected manually. Furthermore due to wind the side lap and
the end lap did not always confirm to the standard configuration
of a photogram-metric block. The approximate values of the
GPS/(INS)-data were not accurate enough to start the process of
automatic tie point generation. This is due to the fact that the
small model planes reveal strong variations in the roll and pitch
angle and also the time synchronisation between the GPS-data
and the images was not always correct. This led to a first step of
manual tie point generation. In the next step automatic tie point
generation could be performed to stabilise the block. Beside
precisely measured ground control points the GPS-positions of
the images were introduced into the aerotriangulation with a
low a priori accuracy of ± 4 m.
Figure 4: Photogrammetric workflow of the UAV-images
The results of the aerotriangulation of the different blocks are
compiled in Table 2.
M 1
M2
W 1
No. of images
17
45
11
Ground resolution (GSD) [m]
0.07
0.08
0.08
No. GCPs
6
20
16
No. of Tie Points
290
1100
384
Total RMS [Pixel]
4.62
1.43
0.38
Residuals
GCP X [m]
0.37
0.14
0.04
GCP Y [m]
0.59
0.08
0.04
M 1
M2
W 1
GCP Z [m]
0.58
0.27
0.03
Image coordinate X [Pixel]
6.44
1.54
0.37
Image coordinate Y [Pixel]
7.75
1.17
0.54
Table 2: Results of the aerotriangulation of the three blocks
The results in Table 2 require further explanation because the
circumstances of the three flights were different. To show the
differences of the two imaging systems the blocks M2 and W1
will be described in more detail.
4.2.1 Flight 2 Merklingsen (M2)
For the block M2 a total of 45 images were selected. Due to the
high speed of the UAV and wind gusts an end lap of 60 %
could not be realised. The side lap is also low or even not
existent between some of the image strips, see Figure 5.
Figure 5: Block Merklingsen 2 (M2)
By using 20 precisely measured ground control points (GCP’s)
the aerotriangulation yielded an inner accuracy of the block
with ± 1.54 pixel (= ± 4.2 pm) in X and ±1.17 pixel (= ± 3.2
pm) in Y, measured in image coordinates. This result is within
the expected limits, taking the block geometry and the
geometric quality of the camera into account. The positional
accuracy on the ground is 0.14 m in X, 0.08 m in Y and 0.27 m
in Z. The accuracy in Z is lower than in X and Y. This is most
probably due to systematic errors in the focal length, a common
problem with consumer grade digital camera with a variable
zoom lens, Remondino and Fraser (2006).
The radiometric quality of the images is quite different due to
different shutter speeds and the fact that some of the images
were true nadir images due to side winds. Even thorough
dodging and colour balancing could not prevent radiometric
differences in the final ortho photo mosaic, see Figure 6.
4.2.2 Block Wahlsdorf (Wl)
The block Wahlsdorf (Wl) consisted of two strips with a total
of 11 images. It has a high side lap (60 %) and an end lap of
80 %. The aerotriangulation with 11 GCPs resulted in a high
internal block accuracy (RMS) of ± 0.36 Pixel (= ± 2 pm) in