You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Proceedings of the Symposium on Global and Environmental Monitoring

relation, if this relation holds for subset S; and does not
hold for the subset S>. The final decision (selection of the
spectral relation for the node) is performed by the
analyst. The tree classifier has several advantages, especially
with multimodal data, because all features are not equally
effective for the description of all target classes.
Finaly, the computer program, which realizes the classification
algorithm, is generated in an automatic way. It is relatively
easy, because the classifier is always of any binary tree
character- A3.1 data necessary for the tree generation are
available in the spectral knowledge data base.
The method described has been successfully used at the Earth
Remote Sensing Centre of the Geodetic and Cartographic
Enterprise in Prague, especially for the classification of
Thematic Mapper data. The main goal was to define the land use.
Several thematic map from the Northeast Bohemia region have
been produced. The best classification result was approximately
92 percent accuracy over 14 target classes (using the
resubstitution estimate of error). The design of binary tree
classifier based on spectral knowledge is rather a time
consuming process. On the other hand, the own classification of
image data does not require a large amount of computing time
and storage.
[ 1 ] Wharton, S. W.: A Spectral-Knowledge-Based Approach for
Urban Land-Cover Discrimination. IEEE Transactions on
Geoscience and Remote Sensing, 1987, No. 3.
[ 2 ] Charvdt, K., Cervenka, V., Soukup, P.: Using Statistical
Tests for Computation of the Classification Parameters in
Remote Sensing of Earth (in Czech). In: Application of
Artificial Intelligence AI 87, DISK, Prague, 1937 .
[ 3 ] Cervenka, V., Charvat, K., Soukup, P.: Automatic
Interpretstion of TM Image Data for Defining the Land Use
(in Czech). In: Application of Earth Remote Sensing Data
in National Economy, CSVTS, Bratislava, 1390.
[ 4 ] Crist, F. P., Ciccne, R. C.: A Physically-Based
Transformation of TM Data - the Tasseled Cap. IEEE
Transaction on Geoscience and Remote Sensing, 1584, No.3.
[ 5 ] Richardson, A. J., Wiegand, C. L.: Distinguishing
Vegetation from Soil Background Information.
Photograrn. Eng. and Remote Sensing, 1 977 , pp. 1541-1552.
[ 6 ) Dusek, D. A., Jackson, R. D., Musick, J. T.: Winter Wheat
Vegetation Indices Calculated from Combinations of Seven
Spectral Bands. Remote Sensing of Environment, 1985, pp.
255 - 267.
[ 7 ] Cervenka, V., Charvat, K.: Nonparametric Classification
Methods in Remote Sensing. In: Application of Artificial
Intelligence AI 90, UISK, Prague, 1990.
[ 8 ] Cervenka, V., Charvdt, K.: Classification of
Multispectrad Imagery Based on Spectral Knowledge £in
Czech). In: Digital Image Processing 89, CSVTS TESLA VUST
A. S. Popova, Prague, 19S9.