
Construction
of the Knowledge Base

The knowledge base is built by means of an interface
, a rule compiler written in PROLOG programming language. The expert expresses
his knowledge in natural language but with predefined keywords related
to
- objects that can be extracted from
GIS:
- valley, slope, ridge, plateau,... related to DEM,
- road, village,... related to topographic map,
- clay, rocky,... related to soils map.
- knowledge description capabilities.
For example, keywords for:
- relational characteristics :
- Far away,
- Near,
- Around,
- Upper ,
- Bottom,
- West, ...
- uncertainty expression :
- Only (+1),
- Principally (+0.8)
- Frequently (+0.5),
- Uncommon (-0.6), ...
Then the interface constructs the corresponding production rules.
Remember that we have two types of rules :
- "pointwise" rules based
on the pixel context itself
- "structural" rules based
on the shape, surface, spatial context of the pixel, etc.


IRIT-UPS