he 
t- 
iC 
  
163 
RECONSTRUCTION OF FREE - FORMED SPATIAL CURVES FROM DIGITAL IMAGES 
Gerald Forkert, Martin Kerschner, Reinhard Prinz and Franz Rottensteiner 
Institute for Photogrammetry and Remote Sensing at the TU Vienna 
GuBhausstrafe 27-28 
1040 Vienna, Austria 
Tel.: +43 1 58801 3802 
Fax: + 43 1 5056268 
e-mail: gforkert@fbgeo1.tuwien.ac.at 
KEY WORDS: curve reconstruction, free-formed curves, edge extraction, bundle block adjustment 
ABSTRACT 
Recent developments in the field of digital image processing suggest to use line information for object reconstruction 
instead of or in addition to traditional methods based on distinct points. We have chosen a car as a real-world test 
object. Images were taken under natural illumination and scanned with a high precision scanner. Point measurement 
for image orientation was supported by image matching techniques. After orientation, edges were extracted from the 
digital images, and the object curves were reconstructed by bundle block adjustment. For that purpose, a new 
concept using free-formed curves was applied. The results show that high accuracy can be achieved by using the 
methods described in this paper. : 
1. INTRODUCTION 
One of the main tasks in photogrammetry is the geometrical reconstruction of 3-D objects from 2-D images. Common 
reconstruction methods are based on distinct points of the object surface. However, it seems more appropriate to use 
edge features in cases where no or hardly any point information is available on the object. Edge features can be 
extracted automatically from digital images using some edge detection algorithm. If homologous edges can be found 
in two or more images, the curve can be reconstructed in object space by spatial intersection of homologous bundles 
of rays. Several methods for the reconstruction of "simple" curves such as straight lines or circles have been 
described in literature (e.g. Zielinski1993). In this work we want to demonstrate the possibility to reconstruct objects 
from digital images using free-formed spatial curves. 
2. THE TEST OBJECT 
z=3m 
£=125"" 
Ss > 
; S 
=3 z=3m 
NS = Fe ¢ 
  
0% LS 
0. 4 
> z=3m e 
£2125" 
Figure 1: car image Figure 2: arrangement of photos 
A car was chosen as a test object for digital object reconstruction. 12 images were taken from that car using a P31 
terrestrial metric camera with a principal distance of 100 mm. The camera was focused to 4m, thus achieving an 
average image scale of 1:40. The image format was 12:cm x 9 cm. Figure 1 shows one of the images; the 
arrangement of the photographs can be seen in figure 2. It was mainly chosen according to considerations about the 
depth of field. For orientation purposes, 116 points were targetted on the car surface using black adhesive paper dots 
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences", Zurich, March 22-24 1995