IMAGE ORIENTATION EXCLUSIVELY BASED ON FREE-FORM TIE CURVES 
Gerald Forkert 
University of Technology Vienna, Institute of Photogrammetry and Remote Sensing 
Gusshausstraße 27-29 / 122 
A - 1040 Vienna, Austria 
Tel.: +43 1 58801 3802 
Fax: + 43 1 5056268 
e-mail: gforkert@fbgeo1.tuwien.ac.at 
Commission Ill, Working Group 2 
KEY WORDS: orientation, reconstruction, bundle block adjustment, digital image, free-form curves 
ABSTRACT 
Recent developments in the field of digital image processing encourage to use line information for image orientation 
and object reconstruction instead of or in addition to traditional methods based on distinct points. In this paper, image 
orientation exclusively based on free-form tie curves shall be demonstrated working on a car as real-world test object. 
Photographs were taken from around the car and scanned in order to obtain digital images. Edges extracted from 
those digital images provided the input data for bundle block adjustment for the determination of both, the 
orientational parameters of the images as well as the spatial shape of the tie curves. The results show, that an 
accuracy in the order of the uncertainty of definition of the object lines could be achieved. 
KURZFASSUNG 
Die jüngsten Entwicklungen auf dem Gebiete der digitalen Bildverarbeitung legen die Verwertung von 
Linieninformation für Bildorientierung und Objektrekonstruktion nahe, um traditionelle, auf Punkten basierende, 
Methoden zu ergànzen oder sogar zu ersetzen. In diesem Beitrag soll mit Hilfe eines Autos als Testobjekt die 
ausschlieBlich auf Freiform-Verknüpfungskurven basierende Bildorientierung demonstriert werden. Mehrere rund um 
dieses Auto angeordnete Photos wurden gescannt, um digitale Bilder zu erhalten. Die von diesen digitalen Bildern 
extrahierten Kanten dienten als Beobachtungsdaten für die Bündelblockausgleichung zur simultanen Bestimmung 
sowohl der Bildorientierungen als auch der ráumlichen Form der Verknüpfungskurven. Die Ergebnisse zeigen, daß 
eine Genauigkeit in der Größenordnung der Definitionsunsicherheit der Objektlinien erzielt werden konnte. 
1. INTRODUCTION 
One of the main tasks in photogrammetry is object 
reconstruction. As a prerequisite the problems of image 
orientation must be solved, which usually is done by 
traditional photogrammetric algorithms based on object 
points. Many objects can hardly be described by points 
but by linear features. It is obvious to use that features 
for image orientation and object reconstruction in 
addition to point based methods. 
Furthermore, line features may be very valuable for the 
automated orientation of digital images: detecting 
homologous lines in two or more images appears to be 
easier in many cases than finding homologous points. 
Several methods have been published which use line 
features represented by a single polynomial (Kager 
1980), what, however, is not sufficient for complex 
object shapes. Spatial free form curves of joined cubic 
polynomials seemed to be appropriate and have been 
implemented into the bundle adjustment program 
ORIENT (Kager 1989). 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 199€ 
The basic concept of this method was presented in 
1993 for the first time (Forkert 1993): The shape of a 
joined cubic polynomial curve is completely described 
by its node points. The positions of these nodes in 
object space are determined by adjusting the curve 
either to object points or to image rays of at least two 
images. 
Applying this method, the three basic problems of 
object reconstruction can be solved using line features 
only: 
1) orientation of at least three images merely by free- 
form tie curves using an arbitrarily chosen local 
reference co-ordinate — system (alike relative 
orientation), 
2) absolute orientation based on free-form control 
curves, and 
3) reconstruction of free-form object curves from at 
least two already oriented images. 
Absolute orientation based on free-form control curves 
was presented in 1994 using models from computer 
196 
    
  
  
  
  
  
  
  
  
   
  
  
  
  
    
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
   
  
  
   
  
  
  
   
  
  
   
  
   
  
  
   
   
    
tomogray 
reconstrt 
car as re 
this pap 
exclusive 
car. 
  
Twelve i 
P31 terr 
of 100 
achievin: 
format v 
images; 
seen in 
consider 
purpose: 
using bl: 
mm. 
In order 
scanned 
PhotoSc 
obtainec