SMALL TO LARGE-SCALE MAPPING USING 
MULTI-RESOLUTION IMAGE DATA 
J. Raggam, A. Almer and M.F. Buchroithner 
Institute for Image Processing and Computer Graphics 
Joanneum Research, Wastiangasse 6, A-8010 Graz, Austria 
Commission IV 
ABSTRACT 
For the international high-mountain remote sensing 
testsite TADAT located in the Austrian Alps near 
Salzburg digital multi-resolution image data have been 
collected covering spaceborne Landsat-TM, SPOT 
and scanned KFA-1000 images as well as airborne 
three-line scanner images and scanned aerial pho- 
tographs. An experiment was initiated to investigate 
the geometric performance of these different data sets 
in the course of ortho-image generation using individ- 
ual images as well as stereo mapping based on im- 
age pairs. The geometric workstation software RSG, 
which is based on state-of-the-art algorithms, has been 
used to generate mapping products in appropriate 
scales from the different input images. Illustration 
and discussion of the basic results is given. 
KEYWORDS: Mapping, Ortho-Image, Quality Con- 
trol, Image Matching, Stereoscopic, 3-D 
1 INTRODUCTION 
The availability of digital remote sensing or pho- 
togrammetric image data in various pixel size opens 
the applicability of such data in cartographic work. 
Typical geoscientific applications are the geocoding of 
digital images, also known as ortho-image generation 
in photogrammetry, or the extraction of 3D informa- 
tion from digital image pairs. 
An international high-mountain remote sensing test- 
site has been established located in the Austrian Alps 
near Salzburg (TADAT, Buchroithner and Kostka, 
1989 [2]). For this, a number of digital image data in 
various resolutions has been collected including space- 
borne Landsat-TM, SPOT and scanned KFA-1000 im- 
ages, but also high-resolution airborne three-line scan- 
ner images and scanned aerial photographs. 
Based on these multi-resolution image data mapping 
applications are presented in this paper concerning 
image geocoding and stereo mapping, together with 
a discussion of the respective results. The follow- 
ing objectives were treated, each of them considering 
parametric mapping models and state-of-the-art algo- 
rithms to achieve high accuracy as possible: 
1. Evaluation of the geometric performance of the 
different image data with reference to the point 
transformation between image and map geome- 
iry. 
2. Geocoding/ortho-image generation of image da- 
ta appropriate to the image pixel size and subse- 
quent quality control of the results. 
3. Evaluation of the geometric performance of stereo 
pairs with reference to the image-to-map stereo 
intersection of projection lines. 
4. Stereo mapping based on automated image corre- 
lation and a selected image pair to derive a digital 
elevation model (DEM). 
All these activities were carried out using the re- 
mote sensing software package RSG (Remote Sensing 
Software Package Graz), which has been developed 
by the Institute for Image Processing and Comput- 
er Graphics (DIBAG) in particular for such geomet- 
ric treatment of remote sensing data. A functional 
overview of this software package is given in Raggam 
et al. (1991, [9]) or GEOSPACE and JOANNEUM 
RESEARCH (1992, [3]). 
2 TEST AREA AND IMAGE DATA 
The TADAT testsite is located some 75 kilometers 
south-east of the City of Salzburg. It contains vari- 
ous topographic features ranging from high mountain 
peaks to a low-lying river valley, and a diversity of 
cultural features covering permanent icefields, culti- 
vated land or built-up areas. For a representative 
part of this test area a DEM is shown in Figure 1 
in an oblique view, which has been derived from topo- 
graphic 1 : 25 000 maps with an initial pixel spacing 
of 25 meters. The following multisensor spaceborne 
and airborne image data have been treated: 
e a Landsat 5 TM image with a nominal pixel size 
of 30 meters (denoted as TM); 
184 
LT, 
    
  
NY 
== 
== 
NN 
   
deri 
The 
airb 
sho