International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
The paper in question reports about Automatic
Aerotriangulation (AAT) within this project which has been
carried out by /ngenieurgemeinschaft Vermessung AVT-ZT
GesmbH (AVT). Existing standard knowledge and rules of the
AAT cannot be transferred to this or similar projects. Hence,
the most interesting and important results and solutions of the
AAT are reported. It will be shown, that the AAT can
successfully be used even in difficult applications such as the
presented project.
2. PROJECT
2.1 Project area, Photo flight, Imagery
In Figure 1 the boarder lines between PAB and the
neighbouring states as well as the other Italian provinces are
documented. The actual project area - part 1 (= tessellated
pattern) - is also visible. Its size is about 220 km.
AUSTRIA
| SWITZER-
LAND Block North
Part 1
Block Middle
Block South [ Reg. Veneto
Prov.
Trentino
S-ATALIA
L— Reg.
Lombardia :
Figure l. Map of the Provincia Autonoma di Bolzano — Alto
Adige with their neighbouring countries, the project-
and block borderlines
By order of the PBA the photo flight of part 1 and 2 where
carried out by Compagnia Generale Ripresearee S.p.A. (CGR).
The flight of part 1 consists of 141 flights lines and 2557
images. It was carried out between July to October 2001. For
this flight, colour negative film and two Wild RC 20 as well as
two Wild RC 30 cameras (focal lens c=15 cm) were used. The
directions of the flight lines were West East. The average photo
scale is 1:13.000. The terrain heights in part 1 reach from 200 -
3.000 m above sea level.
Because of almost vertical rock walls (e.g. in the Dolomites)
and the above mentioned within the project area, there are
height differences of up to 2.000 m within one image. To get
more constant image scales in spite of the mentioned
topographic situations, the flight lines were broken into multiple
parts with different flight heights. Additional flight lines were
performed along the valleys. The average photo scales for these
additional lines are between 1:8.000 — 12.000. Because of the
described configuration, given conditions a ^ very
inhomogeneous image pattern resulted within the blocks. CGR
delivered only approximations of the projection centres. The
declared accuracy of these coordinates was between 0.4 — 1.0
m.
As already mentioned, the duration of four months had been
necessary to complete the flight of part 1. Because of this long
time interval the vegetation between neighbouring. lines may
differ. The mountains are responsible for large shadow parts
within the images produced in September and October. These
facts caused difficulties for the automatic matching process
(e.g. Heipke and Eder, 1998). Similar problems existed for
manual measuring of tie-points too.
CGR was scanning all images using a photogrammetric scanner
ZEISS SCAIL. The images where scanned with a 21 pm
resolution. DVDs were used for data delivery.
2.2 Block division, Control and Check points
Smaller blocks have benefits concerning the photogrammetric
work, the computing time for the adjustments and the error
detection within the adjustment phase. In the project the GIS-
data sets had to be delivered to a given schedule starting from
south to north. Therefore the project area was divided into three
blocks for the AAT- Block South, Block Middle and Block
North. To achieve the required homogeneousness for the AAT
results of the whole block an overlap of four flight lines
between the adjacent sub blocks where used. The result of the
block splitting of part 1 is shown in Figure 1.
The control points have been selected considering the above
mentioned block splitting and the general rules for control
points within aerotriangulation blocks (e.g. Kraus, 2003).
Afterwards the required points were measured using GPS. The
measurements were performed within the GPS-nets of the
Istituto Geografico Militare (I.G.M.) — and the Regione Trento
— Alto Adige. For this determination the given accuracy of the
PAB had to be kept. Existing cadastre points were used as
additional height control points within the aerotriangulation.
3. AEROTRIANGULATION
3.1 Aerotriangulation measurements
Measuring the interior orientation of the images, problems
occurred. A data displacement occurred within some pictures.
These errors were only detected because expected tie points
were missing after the matching process between images of
adjacent lines. Another problem was unexpected high residuals
at the fiducal marks after carrying out the interior orientation of
the images. After informing CGR about this problem, the
images were scanned again. Getting back the new image data,
the images were re-measured. However, the accuracy of the
interior orientation of 171 images was still beyond the expected
values. This problem existed within images taken with each of
the four used cameras. The residuals were between 10 and 30
um, as shown in Table 1.
Number of images Residuals
1 >30 um (»14 Pixel)
20 20 — 30 um (1,0 — 1,4 Pixel)
26 15 — 20 um (0,7 - 1,0 Pixel)
124 10 — 15 um (0,5 - 0,7 Pixel)
171 Total number of images
Table 1. Residuals at the fiducal marks of problematic images
634
In
E
€
B
fi
gt
al
fc
