landscape visualization in high resolution stereoscopic ... · page 36 björn zehner, digital...
TRANSCRIPT
Landscape Visualization in High Resolution Stereoscopic Visualization Environments, Dr. Björn Zehner
Given at the „Digital Design in Landscape Architecture 2008“ Conference in Dessau, Germany, May 2008 [email protected], [email protected]
Page 2 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Motivation
Evaluation of the user’s preferences regarding thePlacement of wind-turbines using choice experiments.Example: What do you prefer for getting a certainAmount of energy ?
20 80m high wind-turbines or 10 120m high ?
Place them in the forest or on open field ?
Page 3 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
UFZ’s Visualization Center
Page 4 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
UFZ’s Visualization Center
Stereoscopic visualizationTracking
Page 5 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Different options to run such a system
Page 6 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
OpenSG against Chromium
Chromium
Pros:+ Transparent for application+ Use of existing applications possible
Cons:- Sending the OpenGL stream from the master to all slave nodes generates high network traffic
OpenSG
Pros:+ After scenegraph has been send to all slave nodes low network traffic
Cons:-Rendering backend of software needs to be reimplemented-Size of scene is limited by main memory of the computer
Page 7 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
UFZ’s Visualization Center
Page 8 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Main contents required
Terrain with texture to represent the ground
Trees in very large numbers (XFrogPlants libraries)
Geometric models, e.g. houses, wind-turbines …
Page 9 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Initial assumption for the workflow
Page 10 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Final workflow
Page 11 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Processing single trees (LOD etc.)
Page 12 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, LOD generation for trunk and branches
Page 13 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, LOD generation for leaves
Page 14 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
With increasing viewer-to-leaf distance the screenspace of the leaves becomessmaller and OpenGL must downsample (filter) the images.
If no series of images is provided this is done automatically,making the leaves more and more transparent.
Page 15 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 16 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 17 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 18 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 19 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 20 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 21 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Tree becomes translucent at a distance
Page 22 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 23 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 24 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 25 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 26 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 27 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 28 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Page 29 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
XFrogTune, mipmapping problem
Tree stays opaque at a distance
Page 30 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Assembling the terrain (ground)
Page 31 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Problems with areal image as ground
2925 x 2775 Meter1625 x 1544 Pixel=> One Pixel ~ 1.8 Meter
In forest regions the floor would show trees from topLow resolution in the near field
Page 32 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Forest with areal image as ground
Page 33 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Forest with perlin noise image as ground
Page 34 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Terrain with image containing usage
color.rgb = code.rgb
Combining differentImages in a shader
Page 35 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Terrain with areal image
color.rgb = (code.r + code.g + code.b + code.a) * areal.rgb
Combining differentImages in a shader
Page 36 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Using a noise image as forest ground
color.rgb = (code.r + code.a) * areal.rgb + (code.g + code.b) * forestground.rgb
Combining differentImages in a shader
Page 37 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Using a noise image as lawn
color.rgb = (code.a) * areal.rgb + (code.g + code.b) * forestground.rgb+ code.r * lawnground.rgb
Combining differentImages in a shader
Page 38 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Adding the geometry
Page 39 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Assembling the scene
Page 40 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Scenegraph-structure for a forest
Page 41 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
From billboards to geometry (shematic)
Billboards(point sprites) Geometry
Page 42 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Scenegraph-structure for a forest (standard quadtree)
Page 43 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Scenegraph-structure for a forest (quadtree with LODs)
LOD nodes
Group nodesEmpty group nodes,trees are renderedas point sprites
Page 44 Björn Zehner, Digital Design in Landscape Architecture 2008, Dessau, Germany
Final scene
~ 20.000 Trees~ 1 GByte~ 15-20 frames/second