IMPACT ON LANDUSE CLASSIFICATION BY GEOMETRICAL PARAMETERS
Heiko Jacobs
Hans Peter-Bahr
Karlsruhe University
Photogrammetry and Remote Sensing
Englerstr. 7
Karlsruhe, Germany
ISPRS Commission VII / Working Group 3
ABSTRACT
Landuse classification is not yet feasible in a general, satisfactory way due to restrictions by many sources.
This paper focuses particularly geometric conditions which influences the final landuse classification result
for Landsat- TM data:
Preprocessing is a somewhat "black box" and often produces data of bad geometric quality which affects the
following procedures.
Minor geometric fidelity of a Landsat scene - e.g. produced by bad control or by insufficient correction for
terrain elevations - will necessarily reduce the quality of the classification result. Using several image sets for
one scene, individual low noise may generate big errors for the final result. In practise this may happen easily,
since landuse classification generally makes sense only in connection with additional georeferenced data.
The individual effects of different geometrical parameters are shown and resulting examples for landuse
classification are displayed.
036