ISPRS, Vol.34, Part 2W2, “Dynamic and Multi-Dimensional GIS”, Bangkok, May 23-25, 2001
282
knowledge representation scheme is employed, consisting of
three components: context, rule base and interpreter. The
inference strategies employs the mixed model combined with
data-driven and model-driven models. In the decision
procedure,
Dempster-Shafer theory is applied for processing all kinds of
information that includes some conflict information. Learning is
one of the most important functions of this subsystem.
Learning is to adapt different situations. Learning results will
be updated into the database.
3.5 Database
Database is the central part which includes not only data
related to environment, resources, land cover, terrain, urban
and other aspects, but also knowledge which is discovered or
undiscovered, stationary or dynamic, commonsense or domain
and so on. On one hand database provide information and/or
knowledge for the above three subsystems, for example,
knowledge for automated feature-based image registration,
feature extraction algorithm and feature connection rule for
automated change feature extraction and identification
subsystem, inference strategies and knowledge base for
Intelligent change feature recognition subsystem. On another
hand, the data and knowledge obtained from these
subsystems will be transmitted into the database. From the
viewpoint of database itself, it is not stationary and invariant
but dynamic. That is said that evolution from data to
knowledge by KDD system always occurs in the inner part of
database. This keeps the freshness of database. Fig 2 shows
database with dynamic behavior in which knowledge is
discovered from data by KDD system.
Knowledge base
1 V
F \
\
\
ledge \
к
KDD System
Informa
>
tion
к
DBMS
' \
Data
Fig.2 Database with dynamic behavior( Data-Information-Knowledge Pyramid )
4.Conclusion
The need for an efficient, accurate, intelligent automated
change information extraction and recognition system from
remotely sensed data will continue to rise. This area of
research will continue to increase in interest as the data
volume becomes larger, data rates become higher, the image
processing ability of machines become faster. Of importance is
to integrate multi-source data, utilize all kinds of knowledge for
better decision and use parallel processing algorithm for
improving the efficiency in the future remote sensing system.
5. References
1. Bruzzone L., Prieto D, Automatic Analysis of the
Difference Image for unsupervised Change Detection,
IEEE Transactions on Geoscience and remote
