ContextDriven Visualization for Augmented Reality - PowerPoint PPT Presentation

1 / 38
About This Presentation
Title:

ContextDriven Visualization for Augmented Reality

Description:

ContextDriven Visualization for Augmented Reality – PowerPoint PPT presentation

Number of Views:165
Avg rating:3.0/5.0
Slides: 39
Provided by: franzl2
Category:

less

Transcript and Presenter's Notes

Title: ContextDriven Visualization for Augmented Reality


1
Context-Driven Visualization for Augmented Reality
  • Dieter Schmalstieg
  • Technische Universität Graz

2
Mobile Augmented Reality
Inertial sensor
camera
HMD
tracked touchpad
GPS
  • Tracking
  • dGPS
  • Inertial
  • Vision (marker)

notebook WLAN GPRS phone
3
Handheld Augmented Reality
  • Low Cost
  • Robust and fool-proof
  • Large variety of devices
  • Intuitive user interface
  • Self-contained operation
  • Networking support
  • Tracking support

4
Example Signpost
  • Wide area 3D navigation system

5
Modeling Requirements for AR
  • AR uses large complex models
  • AR uses geometric, visual models
  • AR does not use normal computer graphics models
  • Visual realism not important
  • Rendering performance not important because only
    limited amount of detail interesting
  • Semantic structure important for application
    goals (situation, context)

6
Overview
  • Problem Data structures and algorithms for
    visualization driven by context
  • Such as location, situation, user prefs etc.
  • Background scene graph
  • Context-driven scene graph traversal
  • Background Magic lenses
  • Context-driven magic lenses

7
Scene Graph
  • Solution for complex graphics problems
  • High-level data structure
  • Objects, not drawings
  • Directed Acyclic graph
  • Hierarchy for
  • Geometric aggregation e.g., articulated body
  • Semantic aggregation, e.g., parts-engine

8
Coins Software Architecture
  • Based on older Open Inventor standard
  • Hierarchical composition
  • Composite pattern
  • Decorator pattern
  • Work done in traversals
  • Visitor pattern
  • Dual-dispatch on node/traversal type
  • Rendering traversal
  • Reflection, garbage collection
  • Scripting language

9
Example
10
Context-driven traversal
  • Problem
  • Information scattered throughout scene graph
  • Finding information requires full search
  • Manipulating may require changing SG in many
    places
  • Acerbated by mix-and-match scene graph
    composition, even at runtime!
  • Need to delegate behavior modification to local
    nodes and their interaction with near nodes

Where on earth is the blue node in the SG?
11
Example
  • Indoor AR navigation
  • 2 views
  • Head up display
  • World in miniature
  • Rendered from same, consistent model
  • No copy!

12
Contribution
  • Generic parameterization during traversal
  • add generic parameter passing technique
  • binding of parameters happens during traversal
  • based on existing standard scene graph Open
    Inventor
  • scene graph becomes a template structure
  • evaluated in the last possible moment

13
Traversal and state
  • State stack
  • Scene graph

1
3
2
  • transformation
  • material parameters
  • light sources
  • state contains specialized information

14
Add general state - context
  • Add 3 things
  • Context state
  • Nodes accessing context
  • Affecting changes in the scene graph

context
15
Operations on context
  • Context maps string gt strings
  • foreground gt 1 0 0
  • position gt 10 -5 0
  • node references
  • Possible operations
  • add
  • set
  • remove
  • clear

16
Implementation
  • Deriving from dedicated classes in Open Inventor
    framework.
  • SoContextElement
  • Stores the context data
  • SoContext
  • Node to write to context
  • SoContextReport
  • Node to read the context
  • SoContextSwitch
  • Reads context and selects children to traverse

17
Setting and retrieving simple context
  • SoContext
  • index foreground,background
  • value 0 1 0, 0 0.5 0
  • DEF REPORT SoContextReport
  • index foreground
  • Material
  • diffuseColor USE REPORT.value

implicit conversion!
18
Templated scene graph
  • Multiple references to one graph
  • Templates for
  • attributes of nodes
  • sub graphs
  • Individual references set template parameters

val
19
Examples mobile augmented reality
world in miniature
augmentation
generic model
20
Examples mobile augmented reality
augmentation
WIM
model
model
wireframe
Z
doorstyle
model
Wall style
R2
R1
generic model
wall
door
21
Discussion
  • Context driven traversal allows local behavior in
    semantic space of SG
  • It does not allow local behavior
  • In 2D screen space
  • In 3D object space
  • ? We want that as well!

22
Example Information Filtering
  • Avoid display clutter for mobile AR user
  • Display only relevant information BOTH
  • Important for user
  • Within field of view
  • Explicitely determined in object space, per
    object
  • We want filtering implicit as part of traversal,
    per pixel

Höllerer 2001
23
Example for geometric locality
  • Subsurface infrastructure visualization
  • (collaboration with Grintec Gesmb)

24
Magic Lenses
  • Local behavior in screen space
  • Bier 1993 2D case
  • Used to
  • reveal hidden information
  • enhance data of interest
  • suppress distracting information
  • Developed in 2D as interaction tools

25
3D Magic Lenses, contd
  • Viega 1996 HW Clipping planes
  • Schmalstieg 1998 Stenciling
  • Flat vs volumetric lenses
  • Flat lenses are a simplification of volumetric
  • Implemented using clipping planes

26
3D Magic Lenses contd
  • Ropinski 2005
  • Uses a more robust multi-pass technique
  • Dual z-buffer by clever use of shadow buffer
  • Programmed as HW shader object space algorithm
  • Single lens only

27
3D Magic Lenses contd
  • Mendez, ISMAR 2006
  • Enhancement to Ropinski 2005
  • Dual-depth buffer using float textures
  • Completely general overlapping lenses

28
3D Magic Lenses contd
  • Mendez, ISMAR 2006
  • Three render passes
  • Render behind and next to lens
  • Render inside lens
  • Render in front and next to lens
  • As many lenses as available float textures but
    render passes

29
Putting It All Together
  • Context-driven Magic Lenses
  • Objects organized in context families
  • Define 1 render style per (lens,context family)
  • Annotate (existing) scene graph with context
  • Targeted as an interaction tool and not as a
    rendering technique

30
Context Families
  • Sub scene graphs (objects) are assigned
    context information
  • Make smart sub graphs who know who they are
  • Correctly styled regardless of their position in
    the graph
  • Styles are also arbitrary sub-graphs
  • Context-sensitive traversal concatenates styles
    and objects in right order
  • Lens nodes create screen/object space masks

Objects affected according to their context
information
31
Building Dynamic Sub Graphs
  • Context-sensitive traversal concatenates styles
    and objects in right order
  • Styling sub graphs are also just concatenated

root content
32
CSML Dynamic Sub Graphs
  • Styling contributions are correctly concatenated
    in case of intersecting lenses (see lt2gt)
  • But no general solution for non-commutative style
    sub graphs yet

lt1gt wrong
lt2gt correct
33
Applications
  • X Ray Vision

34
Applications
  • Information Enhancement

35
Applications
  • Virtual cutout

36
Applications
  • Multiple Magic Lenses

37
(No Transcript)
38
Thanks!
  • Collaborators
  • Gerhard Reitmayr
  • Erick Mendez
  • Denis Kalkofen
  • Michael Knapp
  • Sponsors
  • Austrian Science Fund FWF
  • FFG BRIDGE
  • Grintec GesmbH
  • www.studierstube.org
Write a Comment
User Comments (0)
About PowerShow.com