METHODOLOGY FOR THEMATIC IMAGE PROCESSING USING THEMATIC AND 
the TOPOGRAPHIC DATA BASES AND BASE-INTEGRATED MULTI-SENSOR IMAGERY 
ments for 
for the by 
cation of the ) 
WOLFGANG GÖPFERT 
se map 
letailed 
e Institut für Angewandte Geodásie 
rial: Richard-Strauss-Allee 11 
D-6000 Frankfurt/Main 70 
nall areas Federal Republic of Germany 
ABSTRACT 
For the purpose of thematic image processing it is meaningful to employ 
multi-sensor imagery as input data, supplemented by thematic and topographic 
data bases. 
A method to derive such thematic data bases from existing maps by digit- 
al image processing is presented. 
Due to different sensor modelling functions, different sensor platform 
attitudes at the times of exposure, and topographic relief any acquired imagery 
will be distorted with respect to data base locations. The method proposed here 
differentially corrects the various multi-sensor imagery into the data base 
system, using a rubber-stretch type rectification algorithm. 
Applications are discussed and examples shown. 
1. INTRODUCTION 
For the purpose of thematic classifications and pattern recognitions it 
is meaningful to employ multi-sensor imagery as input data, supplemented by 
thematic and topographic data bases. While different sensors provide additional 
object information in the form of different spectral object signatures, exist- 
ing data bases can be effectively incorporated during the various data process- 
ing steps. 
A topographic data base lends itself conveniently to radiometrically 
correct the imagery for terrain slope effects and to geometrically correct for 
image distortions due to terrain relief. Thematic data bases can be employed 
for the definition and selection of potential training area locations, as well 
as for providing supplementary object informations, such as soil type, geologic, 
hydrologic and other conditions in the image region to be evaluated /1,2,3/. 
A method to derive such thematic data bases from existing maps by digit- 
al image processing is further presented. The approach allows for arbitrary 
input projections of the map and arbitrary output projections for the derived 
data base. 
Due to different sensor modelling functions, and the different sensor 
platform attitudes at the times of exposure any imagery will be always distorted 
with respect to data base locations. Rough topography will further introduce 
distortions necessitating the rectification of the imagery into ortho-imagery. 
The method proposed here differentially corrects the various multi-sensor 
imagery into the data base system, using a rubber-stretch type rectification 
13 
TE Si i