sets with less than 2 matched points, manual selection 
procedure were used. There are total 335 TPs in the 74 sets 
of tie areas. The total number of point measurements are 
1057. They will enter the bundle adjustment with minimum 
constraint. 162 from them were found wrong and discarded 
during the adjustment so that the correctness rate is 8596. 
The posterior sigma naught from the bundle adjustment 
with minimum constraint is 7um. Since the purpose of this 
test is to see the reliability and success rate of the 
matching, we have not done any thing to eliminate 
systematic errors, so that the sigma naught does not 
represent any actual accuracy of the matching. 
7. Conclusion 
As long as only measuring of aerial triangulation is 
concerned, there is no need to digitize the whole 
photograph. For most photogrammetric projects, the 
forward- and side-overlap are very regular as specified in 
the flight contract. In such cases the aerial triangulation 
points always locate at the so called Von Gruber positions. 
Therefore only a very small area needs to be digitized for 
the aerial triangulation. In our case we used the Rollei RS1 
scanner to digitize such small areas for automatic finding 
and measuring the aerial triangulation points. Since the 
data amount of pixels is very small, the whole procedure 
can be done on a PC. The RS1 itself is much cheaper than 
large format high precision whole frame scanners, so that 
our solution can be called as a low-cost solution. 
The only drawback of digitizing only a small area is that it 
has to be precise enough to assure that there will be 
enough overlapping in different photographs. This is 
achieved by manual scanning. Therefore we call this 
solution the semi-automatic measuring of the aerial 
triangulation points as in contrary to the full automatic 
solution like the MATCH-AT from the INPHO Stuttgart or 
the DCCS. 
The test results show that there is no doubt about the 
applicability of this system. In few cases when no point can 
be found at all or the matching is wrong, it always can be 
solved by human intervention. One big problem is that the 
radiometric response of the CCD of the RS1 is not stable 
enough, so that great efforts have been made to reduce 
such errors. 
REFERENCES 
Foerstner, W. 1987: 
A Fast Operator for Detection and Precise Location of 
Distinct Points, Corners and Center of Circular Features. 
Proceedings of the intercommission conference on Fast 
Processing of  Photogrammetric Data. Interlaken, 
Switzerland. pp.281-305 
Helava, U.V. (1987); 
Digital Comparator Correlation System. Proceedings of 
the intercommission Conference on Fast Processing of 
Photogrammetric Data. Interlaken, Switzerland. pp.404- 
418 
Luhmann, T.; Wester-Ebbinghaus, W. (1986): 
Rolleimetric RS - A new system for digital image 
processing. Symposium ISPRS Comm.ll, ISPRS Vol.26/2, 
Baltimore. 
Tsingas, V.: 
Automatisierung | der  Punktuebertragung in der 
Aerotriangulation durch mehrfache digitale Bildzuordnung. 
Dissertation, Deutsche Geodaetische Komminsion, Reihe 
C, Heft Nr.392, Munich, Germany. 
Table 1. Window size and correlation threshold during point matching 
  
  
  
  
  
  
  
Original image First level Second level Third Level 
Target window 35 by. 35 29 by 29 23 by 23 17 by 17 
Search window 47 by 47 41 by 41 35 by 35 64 by 85 
threshold 0.65 0.65 0.65 0.65 
  
  
  
  
Table 2. Number of tie area sets in which 0 to 5 TPs are successfully transferred 
No. of TPs 
transferred 
0 
  
No. of sets 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996