Concepts for Automatic Generalization of Virtual 3D Landscape Models

Transcription

1 Concepts for Automatic Generalization of Virtual 3D Landscape Models Tassilo GLANDER, Matthias TRAPP, Jürgen DÖLLNER Hasso-Plattner-Institut Dep. Computer Graphics Systems Prof. Dr. Jürgen Döllner University of Potsdam

2 Outline 1. Introduction 2. 3D Landscape generalization 3. 3 Generalization techniques for 3D landscape models 1. 3D isocontours of 3D terrain models 2. Cell based generalization of 3D city models 3. 3D generalization lenses 4. Conclusion DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 2

3 1 Introduction Virtual 3D landscape models serve as frameworks for representing geographic and thematic aspects of landscapes. are based on 3D geodata (terrain models, building models, vegetation models, photographic textures). achieve a high degree of realism as required by a number of applications in virtual reality, design, and architecture. [Werner et al. 2005] [Lorenz 2011] DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 3

4 1 Introduction Generalization Fundamental technique (originally in map making) needed for the derivation of a small scale map from a large scale map the creation of user- and task-specific maps Idea: abstract information to present essential aspects while removing unnecessary details, emphasize important aspects and enhance legibility. Aggregation Simplification, Smoothing Selection Enhancement Classification, Typification Displacement [Hake et al. 2002] Exaggeration DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 4

5 2 3D landscape generalization 3D landscape generalization Fundamental operation to transform and present landscape models at a specific and coherent level of abstraction (LOA) Motivated by demands of legibility and comprehension rather than computational requirements Applies, transfers and extends generalization principles from cartography to 3D landscape models Corresponds to typical scale mappings in cartography Requirements for generalization techniques Automated algorithms without human intervention Configurable outcome To provide means for view- and task-specific visualization DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 5

6 2 3D landscape generalization 3D landscape generalization in the visualization pipeline DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 6

7 3 Generalization techniques for 3D landscape models DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 7

8 3 Generalization techniques for 3D landscape models Geometric generalization of terrain models by 3D isocontours Goal: provide a smoothed and de-noised 3D terrain surface while maintaining and enhancing major structures Isocontours (also: isolines, isopleths, level sets) Established visualization for presenting 3D terrain using 2D media Defined by a set of equally spaced isovalues α i Usually mapped to color 1 1 α 2 α 1 0 fmin fmax 0 fmin fmax α DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 8

9 3 Generalization techniques for 3D landscape models 3D isocontours implementation Real-time rendering technique using graphics hardware Per terrain model s triangle: height quantization and tessellation DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 9

10 3 Generalization techniques for 3D landscape models 3D isocontours results Technique independent on terrain origin (TIN/grid) Smooth geometric blending with original terrain Effect control through mask layer [Glander et al. 2010] DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 10

11 3 Generalization techniques for 3D landscape models Cell-based generalization of virtual 3D city models Goal: create abstract city model representation hiding unnecessary details while enhancing infrastructure and landmarks Assumptions Residential buildings are not important Navigation and orientation needs navigable space and landmarks Roads are major structuring elements of a city DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 11

12 3 Generalization techniques for 3D landscape models Implementation Use hierarchical infrastructure network (e.g. roads) to cluster buildings using the formed cells Create building blocks based on cells, extrude to average height Exclude non-building areas (e.g. green spaces, water) Preserve local landmark buildings DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 12

13 3 Generalization techniques for 3D landscape models Cell-based generalization of virtual 3D city models Multi-resolution representations are computed using hierarchy inherent in infrastructure networks (e.g. different road types) [Glander & Döllner 2010] DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 13

14 3 Generalization techniques for 3D landscape models 3D generalization lenses in virtual 3D city models Goal: Emphasize focused parts in the model while hiding complexity in contextual parts Lens metaphor Lenses provide another view to something (e.g., through magnification) Magic, semantic lenses provide in-situ another view to the data, maintaining spatial reference Common metaphor in focus + context visualization DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 14

15 3 Generalization techniques for 3D landscape models 3D generalization lenses in virtual 3D city models Seamless combination of multiple model representations at the same geographic area (LOA representations) Multiple, arbitrarily shaped, overlapping 3D lens shapes Interactive modification (change mapping, lens shape) Implementation Lens shapes are converted into layered depth images (LDI) during preprocessing While rendering LOA geometry, each fragment is tested to be inside or outside the lens shape to decide if it is used DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 15

16 3 Generalization techniques for 3D landscape models 3D generalization lenses in virtual 3D city models Scene lenses (fixed relative to scene objects) 16

17 3 Generalization techniques for 3D landscape models 3D generalization lenses in virtual 3D city models Camera lenses (fixed relative to current view) [Trapp et al.2008] DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 17

18 4 Summary & Conclusion Summary Generalization as a means to reduce comlexity in 3D landscape models Examples have been implemented as research prototypes Conclusions Adaptation of cartographic generalization principles is possible Generalized 3D landscape models as an extension of the visual vocabulary for landscape visualization 3D visualization introduces new challenges to be addressed by generalization Occlusion Free, interactive navigation in the scene Continuous scale in one image, changing scale during navigation DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 18

19 Contact Tassilo Glander Department for Computer Graphic Systems Prof. Dr. Jürgen Döllner Research Group 3D-Geoinformation Download publications at: DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 19

20 References C.A. Brewer & B.P.Buttenfield: Framing guidelines for multi-scale map design using databases at multiple resolutions. Cartography and Geographic Information Science, 34(1):3-15, G.Hake, D. Grünreich, L. Meng: Kartographie, de Gruyter, 2002 H.Lorenz: Texturierung und Visualisierung virtueller 3D-Stadtmodelle, Dissertationsschrift (unveröffentlicht), R.Guercke, T.Götzelmann, C.Brenner & M.Sester : Aggregation of LoD 1 Building Models as an Optimization Problem ISPRS Journal of Photogrammetry and Remote Sensing T. Glander, J. Döllner: Abstract representations for interactive visualization of virtual 3D city models. Computers, Environment and Urban Systems, 33(5): , T.Glander, M.Trapp & J.Döllner: 3D Isocontours Real-time Generation and Visualization of 3D Stepped Terrain Models. Eurographics 2010 Shortpaper, 17-20, M.Trapp, T.Glander, H.Buchholz & J.Döllner: 3D Generalization Lenses for Interactive Focus +Context Visualization of Virtual City Models. 12th International Conference on IEEE Information Visualization, pp , A. Werner, O. Deussen, J. Döllner, H.-C. Hege, P. Paar and J. Rekittke: Lenné 3D walking through landscape plan. In: E. Buhmann, P. Paar, I. Bishop, and E. Lange, editors, Trends in Real-Time Landscape Visualization and Participation, pp Wichmann Verlag, DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 20

21 How to cope with 3D terrain models in cell based generalization Naive approach: tessellate add height offset per vertex Better: use height difference as a condition in aggregation See [Guercke et al. 2011] DLA T.Glander: Concepts for Automatic Generalization of Virtual 3D Landscape Models 21