Finding topographically-similar regions in a triangulated terrain model
Two shortcomings of systems developed to automatically extract the topographic features of terrain are that feature definitions are often based on a single attribute, such as Gaussian curvature or slope, and that interactive queries regarding the similarity of one region to another are rarely sup...
| Main Author: | |
|---|---|
| Language: | English |
| Published: |
2009
|
| Online Access: | http://hdl.handle.net/2429/5945 |
| Summary: | Two shortcomings of systems developed to automatically extract the topographic features of
terrain are that feature definitions are often based on a single attribute, such as Gaussian
curvature or slope, and that interactive queries regarding the similarity of one region to another
are rarely supported.
We report on a prototype system for triangulated terrain models that attempts to address these
two issues. It supports interactive classification of the terrain via a user-controlled classification
system. Each classification category is defined in terms of ranges of five key terrain attributes:
local relief, slope aspect, slope gradient, plan curvature and profile curvature. The system
also supports queries of the type: "Which regions are topographically-similar to the selected
region?" We use optimal-monomorphism in attributed graphs so that "topographically-similar"
goes beyond simply matching terrain attributes to include the attributes and spatial relations
of adjacent regions. |
|---|