Full text: XVIIIth Congress (Part B3)

e From the 186 perfect image triples 149 succeed 
when submitted to the iterative registration procedure. 
There are 37 constellations which yield a final number 
of hits in the range of 10 to 45 which is again signifi- 
cantly lower than the acceptance threshold. Note, that 
a set of 45 correspondences implies that there are at 
least 38 correct matches which were missed due to too 
large residuals. Thus, the resulting orientation esti- 
mate cannot be satisfying and, consequently, rejecting 
a total match which does not yield the maximum num- 
ber of 83 correspondences is appropriate and justified. 
e The number of 149 successfully matched image triples 
represents 7096 of all perfect constellations and 2296 
of the total number of valid image triples. Thus, we 
can conclude that on average we have to perform 4.5 
random trials until the matching process signals suc- 
cess. All inappropriate constellation matches are guar- 
anteed to be rejected, while the choice of the accep- 
tance threshold is highly uncritical. 
4. SUMMARY AND CONCLUSION 
We presented an approach for automatic exterior orientation 
of aerial images which is based on the use of manhole cov- 
ers as landmarks. For this purpose we developed a landmark 
extraction scheme which enables us to automatically detect 
a quarter up to a third of all manhole covers visible within 
an image while the error rate is reasonably low. Using the 
positions of the detected landmarks for estimating the ori- 
entation parameters requires to identify each landmark, i.e., 
to find the correct correspondence with a landmark from the 
cadastral database. This database includes the geodetic co- 
ordinates of all manhole covers that exist within the observed 
area. Thus, we have to match landmark constellations, where 
the set of image landmarks is incomplete and distorted (un- 
detectable covers and false detections, resp.). We introduced 
a matching approach which is based on matches of triples of 
landmarks. In a randomized selection process a triple match 
is hypothesized; the match is verified by estimating the orien- 
tation parameters and looking for additional correspondences, 
both combined in an iterative process. Candidate selection is 
efficiently implemented using indexing through geometric in- 
variants. In our experiments the matching approach proofed 
to be successful: It robustly yields the correct total match. 
Based on a large number of correspondences, the final match 
provides a reliable estimate of the exterior orientation param- 
eters. 
We conclude that by combining both, a robust and precise 
landmark extraction scheme with a reliable matching mech- 
anism, we are able to provide an effective approach to au- 
tomatic exterior orientation of aerial images. Our approach 
offers the opportunity to automate the orientation process 
in urban environments, presuming that the approximate im- 
age scale and the approximate location of acquisition can be 
derived from the flight plan. It requires that the geodetic 
coordinates of manhole covers are available from cadastral 
databases in digital format. This might not be the typical 
case today, but triggered by the transition from analog to 
digital photogrammetry and by the rapidly spreading use of 
GIS-technology, there is a strong trend towards digitalization 
in many cadastral tasks. 
152 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996 
SOURCES OF DATA 
Aerial photographs and map material used in our experiments 
have been kindly provided by the Vermessungsamt Hamburg. 
For illustrations we used the image FIR 87-20-828 acquired 
by the Gesellschaft fiir technische Photogrammetrie m.b.H, 
Hamburg. Reproduction is done with the permission of the 
Vermessungsamt Hamburg. Digital coordinates extracted 
from the cadaster of the sewerage system of the city of Ham- 
burg have been kindly provided by the Hamburger Stadtent- 
wasserung, Anstalt des Offentlichen Rechts, Hamburg. 
ACKNOWLEDGMENTS 
We thank Rainer Sprengel for constructive suggestions with 
respect to the landmark model, Joachim Rieger for suggest- 
ing the use of invariants, and Carsten Schroder for support 
in the implementation and for making valuable contribu- 
tions to the revision of this manuscript. For providing the 
data material we thank Mr. Rohwedder (Vermessungsamt), 
Mr. Sachweh (Vermessungsamt), and Mr. Gleim (Stadtent- 
wasserung). Parts of the work reported here have been done 
in the LUKAS-project carried out at the Al-Laboratory at 
the Computer Science Department, University of Hamburg, 
in cooperation with the Ingenieurbüro Basedow and Tornow 
GmbH, Hamburg. 
REFERENCES 
Drewniok, C. and K. Rohr, 1995. High-Precision Localiza- 
tion of Circular Landmarks in Aerial Images. In Sagerer, G., 
S. Posch, and F. Kummert, editors, Mustererkennung 1995 — 
17. DAGM-Symposium, Bielefeld, Germany, 13.-15. Septem- 
ber 1995, Informatik aktuell, pages 594—601. Springer-Verlag, 
Berlin — Heidelberg — New York, 1995. 
Gülch, E., 1995. From Control Points to Control Structures 
for Absolute Orientation and Aerial Triangulation in Digital 
Photogrammetry. Zeitschrift f. Photogr. u. Fernerkundung 
63 (3), 130-136, May 1995. 
Lenz, R. and P. Meer, 1994. Point Configuration Invariants 
Under Simultaneous Projective and Permutation Transforma- 
tions. Pattern Recognition 27 (11), 1523-1532, 1994. 
Meer, P., S. Ramakrishna, and R. Lenz, 1993. Correspon- 
dence of Coplanar Features Through P?-Invariant Represen- 
tations. In Mundy, J.L., A. Zisserman, and D.Forsyth, editors, 
Applications of Invariance in Computer Vision, Ponta Del- 
gada, Azores, Portugal, October 1993, volume 825 of Lecture 
Notes in Computer Science, pages 473-492. Springer-Verlag, 
Berlin — Heidelberg — New York, 1994. 
Schickler, W., 1992. Feature Matching for Outer Orienta- 
tion of Single Images Using 3-D Wireframe Controlpoints. In 
Proc. XVII ISPRS Congress, Washington, D.C., 1992, volume 
XXIX, part B3 of International Archives of Photogrammetry 
and Remote Sensing, pages 591-598. American Society for 
Photogrammetry and Remote Sensing, Bethesda, MD, 1992. 
Schickler, W., 1995. Ein operationelles Verfahren zur au- 
tomatischen inneren Orientierung von Luftbildern. Zeitschrift 
f. Photogr. u. Fernerkundung 63 (3), 115-122, May 1995. 
Vosselman, G. and N. Haala, 1992. Erkennung topographi- 
scher PaBpunkte durch relationale Zuordnung. Zeitschrift 
f. Photogr. u. Fernerkundung 60 (6), 170-176, Nov. 1992. 
    
  
   
  
   
    
   
    
   
   
   
   
   
    
   
       
  
    
   
   
   
  
    
   
   
    
    
     
   
   
    
   
    
     
     
    
     
     
    
    
   
   
   
   
   
   
    
KEY WORF 
ABSTRAC 
Convention 
coordinates 
along the p 
coordinate 
control the 
project is tl 
Global Pos; 
of man mac 
different G 
adjustment 
Since the 
satellites in 
application 
triangulatic 
at the ir 
determined 
combined ( 
observatior 
control poi 
completed 
signals are 
and as long 
Howevet, t 
assisted ae 
calibration 
ambiguity 
(Ackermar 
After kinei 
the antenn 
frame. Mo 
national c 
between tl 
formulas (! 
available i 
ellipsoid a
	        
Waiting...

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.