Transcript notes - Virtual Globe

Applied Mathematics
Distributed visualization
of terrain models
How to get the whole world
into a coffee mug...
Rune Aasgaard
1
Applied Mathematics
Where to put the workload?
 Do everything at the server
 Requires a powerful server...
 …and fast network connection...
 ...but simple client.
 Render in the client
 Reduces load on server and network…
 …smooth interactive movement actually possible…
 …but requires a smart and complex client...
 …and more sophisticated hardware.
2
Applied Mathematics
Where to put the data?
 Client terrain database
 Near graphics system
 Fast updating from server data
 Limited size
 Some support for simple analysis
 Server terrain database
 Huge data volume
 Fast query access
 No traversal of data
 Integration of new and improved data sets?
3
Applied Mathematics
Level-of-Detail Triangulation
 Consists of:
 A coarse base triangulation: T0
 A set of refinement operations: Ti
 Results in:
 A set of triangulations: Ti
 View dependent expansion of client data structures:
 Only show what is necessary for generating an image
 Use screen-space error tolerance
 Approximation error estimates for each refinement operation
4
Applied Mathematics
Client data structures
Should support the graphics system





Triangle strips
3D coordinates
Surface normals
Texture coordinates
Map to a set of texture tiles
 Portability - Java and Java3D
5
Applied Mathematics
Client data structures
Update with data from server
 Start with coarse base triangulation
 Request data from server when:
 Area becomes visible
 More detail is required (viewpoint moved in)
 Reduce to coarser level when:
 Area becomes invisible
 Less detail is required (viewpoint moved out)
6
Applied Mathematics
Server data structures
Can be huge!
 Whole earth, 30” grid (DTED Level 0): 933.120.000 points!
 Whole earth, 3” grid (DTED Level 1): 93.312.000.000
points!
 Luckily, 2/3 of the earth is ocean
 Major parts of the land is relatively flat
 Can benefit from data simplification and compression
7
Applied Mathematics
Server data structures
 Server responds to client requests:
 in: Position
 out: Elevation and Elevation approximation error
 Queries are expected to be:
 chunked
 localized in area and resolution level
8
Applied Mathematics
Binary Triangle Trees
 Hierarchy of right-isosceles triangles
 Related to Lindstrom triangulations and the ROAM
algorithm
9
Applied Mathematics
Binary Triangle Trees
 Simple data structures
 simplifies network streaming
 Regular refinement pattern
 fits well with texture tiles
 simple integer coordinates
 maps easily to regular quad trees
 But….
 requires more triangles for representing complex objects than
irregular triangulations
10
Applied Mathematics
Approximation error spheres
 One sphere for each
vertex
 Radius =
Approximation error
/ angular resolution
 If the viewpoint is
inside sphere,
display vertex
11
Applied Mathematics
Zooming in - Scandinavia
12
Applied Mathematics
Zooming in - Scandinavia
13
Applied Mathematics
Zooming in - The Oslo fjord
14
Applied Mathematics
Zooming in - The Oslo fjord
15
Applied Mathematics
Zooming in - Tønsberg
16
Applied Mathematics
Zooming in - Tønsberg
17
Applied Mathematics
San Francisco - bay area
18
Applied Mathematics
Islands in the sun
19
Applied Mathematics
Oslo fjord - elevation color coding
20
Applied Mathematics
Oslo fjord - elevation color coding
21