PHOTOINTERPRETATION, UNSUPERVISED OR SUPERVISED CLASSIFICATION ?
FRIEDRICH QUIEL
Institut für Photogrammetrie und Topographie
Universität Karlsruhe
ENGLERSTRASSE 7
7500 KARLSRUHE GERMANY
ABSTRACT
Remote sensing data contain a wealth of information and a variety of evaluation tech-
niques are used. The most suitable procedure should be selected to extract the information
necessary to solve a specific problem. To reach this goal, the requirements of the
application and the possibilities of the evaluation procedure should match as close as
possible,
Different sets of requirements are discussed and examples in geology and land use mapping
are used to demonstrate the possibilities and limitations of photointerpretation, unsuper-
vised and supervised classification.
Common problems in photo interpretation and classification are feature selection,
determination of characteristic values for classes, need for ground information and
incorporation of additional information.
Classifications are mainly restricted to spectral and/or textural classes. The additional
use of size, shape, pattern etc. would expand the application of classification techniques
to other problems and data.
The combination of photointerpretation and classification is a very powerful tool to
economically meet the requirements for many problems. It can be supported very
effectively by the use of different data, e.g. LANDSAT and aerial photographs.
INTRODUCTION
With the availability of digital remote sensing data, computer supported digital
evaluation techniques become increasingly important. In using remotely sensed data to
solve a specific problem, a decision must be made, which of the different evaluation
techniques and data are most suitable for the task at hand. It is therefore appropriate to
discuss the advantages, disadvantages, requirements and limitations of the major techniques,
i.e. photointerpretation, supervised and unsupervised classification in general terms.