A new technique is presented for visualizing high-resolution terrain elevation data. It produces
realistic images at small scales on the order of the data resolution and works particularly well when
natural objects are present. Better visualization at small scales opens up new applications, like site
surveillance for security and Google Earth-type local search and exploration tasks that are now done
with 2-D maps. The large 3-D maps are a natural for high-resolution stereo display.
The traditional technique drapes a continuous surface over the regularly spaced elevation values.
This technique works well when displaying large areas or in cities with large buildings, but falls
apart at small scales or for natural objects like trees. The new technique visualizes the terrain as a
set of disjoint square patches. It is combined with an algorithm that identifies smooth areas within
the scene. Where the terrain is smooth, such as in grassy areas, roads, parking lots and rooftops, it
warps the patches to create a smooth surface. For trees or shrubs or other areas where objects are
under-sampled, however, the patches are left disjoint. This has the disadvantage of leaving gaps in
the data, but the human mind is very adept at filling in this missing information. It has the strong
advantage of making natural terrain look realistic, trees and bushes look stylized but still look
natural and are easy to interpret. Also, it does not add artifacts to the map, like filling in blank
vertical walls where there are alcoves and other structure and extending bridges and overpasses
down to the ground.
The new technique is illustrated using very large 1-m resolution 3-D maps from the Rapid Terrain
Visualization (RTV) program, and comparisons are made with traditional visualizations using these
maps.
|