Smooth, Interactive Rendering and On-line Modification of Large-Scale, Geospatial Data in Flood Visualisations
•
0 likes•448 views
This document discusses techniques for interactively rendering and modifying large-scale geospatial LiDAR point set data. It proposes a rendering-on-budget approach that combines importance-based streaming with a PID controller to balance load. This allows for smooth rendering while modifying streamed data online without quality loss. Proof of concepts demonstrate modifying attributes like color via polygons or textures, and displacing vertices using displacement maps. Performance is improved over traditional level-of-detail approaches.
Report
Share
Smooth, Interactive Rendering and On-line Modification of Large-Scale, Geospatial Data in Flood Visualisations
- 1. 1Challenge the future Smooth, Interactive Rendering and On-line Modification of Large-Scale, Geospatial Data in Flood Visualisations
- 2. 2Challenge the future Introduction • 3D Geospatial Data very large and heterogeneous • applicable in e.g. hydrology and climatology Cloud Vis of KNMI dataWater Vis of 3Di subgrid data
- 3. 3Challenge the future Introduction • topographic data nowadays captured via LiDAR • AHN-2 (Actueel Hoogtebestand Nederland), coloured: ~14 TB • data too big for rendering • interactive, on-line modification not possible without quality loss
- 4. 4Challenge the future Approaches and Challenges • Rendering Large-Scale data: Out-of-Core LoD structure (see Kehl et al. ICT Open 2012 for starting point) • Issue: How do modify streamed data, not being constantly available ? • Modification algorithm needs to handle detail-varying data • idea: modify what you see on-chip modification LoD’s 0 1 2 3
- 5. 5Challenge the future • traditional LoD: visual jumps when loading new buckets [1] • solution: Rendering-on-budget for LiDAR point sets • combined importance-based streaming (similar to Sequential Point Trees [2]) with PID controller for load balancing Rendering-on-Budget Methods [1] Christian Kehl and Gerwin de Haan. Interactive simulation and visualisation of realistic flooding scenarios. In Intelligent Systems for Crisis Management, 2012. [2] Carsten Dachsbacher, Christian Vogelsang, and Marc Stamminger. Sequential point trees. In ACM Transaction on Graphics, pages 657-662, 2003.
- 6. 6Challenge the future • Interface to Geo-Information: GoogleMaps KML polygons • conversion from polygons to triangular mesh via constrained DT • exclusion from exterior triangles via polygonal restriction • storage of triangular mesh and attributes in Quadtree • storage of Quadtree in GPU Texture • on-the-fly evaluation of Quadtree per vertex on GPU during rendering • application of attribute modification based on triangle data On-line Modification of Large-Scale, Geospatial LiDAR point sets Methods
- 7. 7Challenge the future On-Line Modification of Large-Scale, Geospatial LiDAR point sets Methods
- 8. 8Challenge the future • Attribute modification possibilities: • colour via pre-defined polygon colour (RGBA) • vertex rendering discard via polygonal area • colour via painting on texture for polygon • displace vertices via painting on displacement map • Also possible to adapt paths (line segments along streets) given via GoogleMaps On-Line Modification of Large-Scale, Geospatial LiDAR point sets Methods
- 9. 9Challenge the future proof of concept – colour via polygon Results
- 10. 10Challenge the future proof of concept – colour via texture Results
- 11. 11Challenge the future proof of concept – displace via texture Results
- 12. 12Challenge the future proof of concept – real-world scenarios Results today 1953 dyke adaptation
- 13. 13Challenge the future performance measurements Results Comparison of rendering behaviour of initial approach (left) and Rendering-on-Budget (right)
- 14. 14Challenge the future performance measurements Results