intervention for tie point measurement. The robust block ad-
the images show wooded, rural or city areas.
justments carried out by PAT-B eliminated approximately 4%
of the tie points determined automatically, regardless whether
PHODIS AT currently needs approximately 6 minutes for
measuring an image on a Silicon Graphics Workstation with
R4400 processor and 200 MHz.
Since the time for the digital aerotriangulation is derived, 20%
error margin is added. Despite this addition, great time differ-
ence can be found. Considering that much of the given time is
pure computing time (eg. image scanning with autowinder,
pyramid generation, automatic interior orientation, automatic
tie point determination) by the digital aerotriangulation meas-
urement, it becomes even more obvious how economic the
digital aerotriangulation by PHODIS AT is.
Block preparation and measurement for 659 images:
Table 4
Hours
30
Area cation 105
Point transfer 600
Data ent 25
Measurement (* 1200 (220
Total 1960 (980
* The 1200 hours result from special demands. In normal
cases, one can expect to measure up to 3 images per hour
which results in 220 hours.
Table 3
Forssa | Moeck | Graz | Echallens
No. of Images 28 15 12 9
Resolution 15 15 28 20
Oo (um) 6.4 6.9 8.4 4.1
Co (um) 3-3 7 XD)
(analytisch) 23.0(E)
Image scale 4000 7000 4000 5500
Focal length 153 153 305 153
No. of GCPs SS 7 36 10
No. of tie points
6-fold 31 19 4 11
5-fold 49 48 26 20
4-fold 228 133 102 68
3-fold 1238 1003 819 434
2-fold 3732 1862 2701 1509
Image 2.42 2,8] 3.35 1.89
rms x/y (um) 3.74 3.92 5.62 2.68
Object 0.015 | 0.016 0.028 0.013
rms x/y/z 0.016 | 0.012 0.025 0.012
(m) 0.022 | 0.018 0.047 0.015
P = planimetry; E = elevation (computed by PAT-M)
When considering the adjustment results, one can focus on os.
o, indicates the accuracy of the automatically measured points
and can be used for comparision with that of an analytical
aerotriangulation. The analytically determined values given in
Table 3 were obtained at those points marked by a point trans-
fer instrument. The accuracy of digital aerotriangulation
amounts to ca. 0.2 to 0.3 pixel size. The digitally determined
exterior orientations are clearly more reliable than that of the
analytically determined ones because they are obtained with a
much heigher redundancy. It shows in Table 3 that the larger
the overlap is (eg. block Echallens), the better the block ad-
justment results, which has already been known for a long time
from the block adjustment theory.
Our tests have shown that the accuracy of block adjustment
does not decrease linearly with the resolution of the images.
Halfing the image resolution (e.g. 30um instead of 151m)
results in an accuracy loss of ca. '^4. This can be explained by
the large number of points, which ensures that the result is
very stable despite the reduced resolution. This is true for good
image material. If the quality of the image material is poor, the
loss of accuracy can be larger, which can be shown by the
block Forssa. An adjustment accuracy of o, — 11.9 was ob-
tained for this block with an image resolution of 30 um.
For the block Graz we have detailed information on the time
and costs for an analytical aerotriangulation (Ganster/Xu,
1994). If one considers only the time for analytical measure-
ment (Table 4) and compares it with the derived time for the
digital aerotriangulation conducted by PHODIS AT (Table 5),
it becomes clear how economic the digital aerotriangulation
can be.
Table 5
Digitally Hours
Scanning 165 (15 min/image)
Data management 220 (20 min/image)
Block definition 8
Pyramids computation 33 (3min/image)
Automatic interior orientation 11 (1 min/image)
Automatic measurement 66 (6 min/image)
GCP measurement & adjust- 60 (4 min/point)
ment (**)
Total 555 + 20% = 666
(**) In the block there are 909 GCPs. Four minutes per GCP
are assumed. This is very liberal especially when considering
that PHODIS AT can guide operator to the GCPs after a first
block adjustment. System-guided GCP measurement does not
require more than 2 minutes per point.
7 Conclusion and Outlook
PHODIS AT is a new system available on the market that
performs fully automatic tie point determination and semiau-
tomatic control point measurement for aerotriangulation.
PHODIS AT is already proving itself in practical use. The high
degree of automation makes aerotriangulation more economic.
The vision is near at hand that a chain can be build up which
allows automatic image scanning, automatic aerotriangulation,
automatic DEM and orthoimage generation, and even auto-
matic stereoscopic or monoscopic data aquisition for geo-
graphic information systems. These prospects are very promis-
ing and can be achieved already in the near future.
References:
Ackermann, F., 1983. High Precision Digital Image Correlation.
Schriftenreihe des Instituts für Photogrammetrie der Universität Stuttgart,
Heft 9, 231-243.
Ackermann, F., 1995. Automatic Aerotriangulation. In Proceedings 2nd
Course in Digital Photogrammetry, Bonn.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996
Ackeı
tion. /
Ganst
Magi:
Gülct
grami
IfP, 1
(IfP),
Autor
Stuttg
Mayr.
gramt
Schen
Photo
Scher
Systei
Tang,
Geoir
Tang,
tation
Wichi
Tang,
Image
47-55
Tang,
matic
accep
Toth,
Autor
tion, [
Tsing
Aeria
214.
Tsing
angul:
Münc
