156
as a gross error (10 mm). Due to the erroneous observation it received such a weight that it showed the characteristics of a leverage
point (ref. chapter 4.). This effect is eliminated by balancing.
The newly developed methods mentioned in this paper have been implemented in the program NAWE of the RolleiMetric
software. NAWE computes initial values of object co-ordinates and parameters of exterior orientation for the bundle adjustment.
The modified NAWE was then compared with the original in calculation of large image bundles in order to document the increased
efficiency in image orientation by the results of this paper.
The determination of approximate values was three times faster with the new NAWE than it was with the old version for all image
bundles. Furthermore, about 10% more object points were determined in all examples. A main reason is the reliable error detection
with the robust estimation procedure. In the determination of approximate values based on the L;-norm-estimation, observations
are deleted which do not show any error.
6. PERSPECTIVES
The results of this paper show the advantages of the introduction of balanced adjustment which make the introduction of this
concept also relevant in other fields of photogrammetry and digital image processing. An example is the bundle adjustment. By
using balanced adjustment, efficiency and reliability of the results will be significantly increased . The combined use of different
Ly-norm-estimation procedures can furthermore be taken into consideration (ref. Kampmann 1988). The balanced L;-norm-method
is used for the error detection. After having eliminated all blunders, a parameter estimation following the L,-norm or the minimax
method corresponding to the aim are appropriate methods depending on the problem to be solved. After having successfully
implemented the new evaluation technique into the program NAWE, the above mentioned problem of the bundle adjustment is the
subject of a joint project of Rollei Fototechnic and the Ingenieurgesellschaft Dr. Kampmann.
There are new perspectives in the multi-image triangulation. Calculations have become faster and more reliable which reduces
costs. The availability of reliable orientation programs is necessary in many applications in close range photogrammetry. One
example e. g. is the industrial photogrammetry. For the method of multi-image triangulation there are applications in car
manufacturing, aerospace and shipbuilding industry. In many cases a further usage is restricted by the long time necessary for the
orientation calculation. With the results presented in this paper, the possibility is given to introduce photogrammetry as a standard
technique in industry and to increase the number of applications.
7. LITERATURE
Fellbaum M. 1994: Robuste Bildorientierung in der Nahbereichsphotogrammetrie. Geodátische Schriftenreihe der
Technischen Universitát Braunschweig, Nr. 13
Grafarend E., Heister H., Kelm R., Kropff H., Schaffrin B. 1979: Optimierung geodätischer MeBoperationen. Verlag Herbert
Wichmann, Karlsruhe
Kampmann G. 1988: Zur Kombinativen Norm-Schétzung mit Hilfe der L;-, der L,- und der Boskovic-Laplace-Methode
mit den Mitteln der linearen Programmierung. Veröffentlichung des Geodätischen Institutes der Rheinisch-
Westfälischen Technischen Hochschule Aachen, Nr 43
Kampmann G., Wolf H. 1989: L1-Norm-Schátzung mit korrelierten Beobachtungen. Zeitschrift für Vermessungswesen,
No. 10, pp. 493-497
Müller H. 1986: Zur Berücksichtigung der Zuverlàssigkeit bei der Gewichtsoptimierung geodátischer Netze. Zeitschrift für
Vermessungswesen, No. 4, pp. 157-169
Rousseeuw S.J., Leroy M. A. 1987: Robust Regression and Outlier Detection. John Wiley and Sons, New York
Wester-Ebbinghaus W. 1985: Bündeltriangulation mit gemeinsamer Ausgleichung photogrammetrischer und geodátischer
Beobachtungen. Zeitschrift für Vermessungswesen, No. 3 , pp 101-111
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop “Crom Pixals to Sequences”, Zurich, March 22-24 1995