a method for creating papercraft raised relief maps from ......complexity in small-scale relief...
TRANSCRIPT
A Method for Creating
Papercraft Raised Relief Maps
from Digital Elevation Models
Jürnjakob Dugge, [email protected]
Johann Dugge, [email protected]
http://www.PapercraftMountains.com
Motivation
One-off physical representations of elevation models
Wolfgang Knoll, Martin Hechinger
Architectural Models: Construction Techniques.
Munich: Deutsche Verlags-Anstalt, 2006
Low-poly aesthetic
Timothy J. Reynolds
Map Wars
2013
Timothy J. Reynolds
Untitled
2014
www.turnislefthome.com
From digital elevation model
to physical paper model
Chil / Zacharie Grossen, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=16791896
Process Overview
Process Overview
Mesh Generation
Greedy Insertion [1]
Sliver triangles
Greedy Insertion
with triangle shape constraints [2] [3]
Flipped edges
Mesh Optimisation Method
Mesh Optimisation Method
1. Objective Function
2. Mesh Generation & Optimisation
3. Results
Objective Function
DEM Feature Weighting
DEM Approximation
Triangle Shape
Objective Function
DEM Feature Weighting
Chil / Zacharie Grossen, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=16791896
• All grid points have weight 𝑤 ≥ 1
• Important geographic features / areas should be prioritised, 𝑤𝑖 ≫ 𝑤
Objective Function
DEM Approximation
• Minimise sum shortest distance to underlying triangle squared, weighted
• Normalised to sum of weights, and weighted DEM variance
𝑀𝐷 = 1 −1
𝑉𝑎𝑟(𝑤𝐷)
1
𝑖=1𝑛 𝑤𝑖
𝑖=1
𝑛
𝑤𝑖𝑑𝑖2
• 3D shape (not projected 2D) considered
• 𝑀𝑆 ="Mean Ratio Metric” target shape equilateral triangle [4]
• Maximise mean of triangles
Objective Function
Triangle Shape
Objective Function
Objective Weighting
Balance optimisation between deviation reduction and triangle shape
𝑀𝑇 = 𝑐𝐷𝑀𝐷 + 𝑐𝑆𝑀𝑆
𝑐𝐷 : 𝑐𝑆 1 : 0 1 : 1 0 : 1
𝑀𝑇 0.9717 0.9195 0.9057
𝑀𝐷 0.6348 0.9239 0.8369
𝑀𝑆 0.9717 0.9151 0.9057
Mesh Generation & Optimisation
DEM Feature Weight Map Generation
Vertex Placement
Quad Flipping
Vertex Shifting
Shape Heuristic
Objective Function Gradient
Mesh Generation & Optimisation
DEM Feature Weightmap Generation
• Identify areas of high local (planar) curvature [5]
• Gaussian blur to remove small scale features
Mesh Generation & Optimisation
Vertex Placement
1. Given target number of triangles n, determine number of vertices inside Pin / on
boundary Pb.
2. Total DEM Area / n ideal equilateral triangle edge length into DEM boundary length
Pb
Pin = round ( 1/2 * (n - Pb + 2) )
Pb = n – 2 * Pin + 2
3. Vertices into corner, interpolate surface
4. Next point at highest deviation * weight * discountAnimation:
https://youtu.be/3xSo2XK8UQY
Mesh Generation & Optimisation
Vertex Placement Examples
Nearest
Linear
v4
Animation:
https://youtu.be/3xSo2XK8UQY
Animation:
https://youtu.be/Lwu31t1GTTA
Animation:
https://youtu.be/wtMNen0WD2U
Mesh Generation & Optimisation
Quad Flipping
Delaunay Triangulation
Identify convex quads
Rank quads by contribution to objective total MT
Flip (next) lowest ranking quad finish when all quads tested
Contribution to MT improved?
yes
no
Mesh Generation & Optimisation
Vertex Shifting
yes
Pull vertices towards
deviation centroids
Gravity / Spring Heuristic
Pull vertices by edge
springs
TM
improved?
Steepest Ascent
Move points in direction of best
TM improvement
TM
improved?Increase step size
Undo move
Decrease step size
yes
Quad Flipping
Quad Flipping
Mesh Generation & Optimisation
Vertex Shifting – Gravity / Spring HeuristicStable, all triangles guaranteed to retain
orientation.
Large steps possible especially for shape
improvement.
Deviation
Centroid of weighted deviation for each
triangle.
Deviation centroids pull each point attached to
triangle.
Shape
Model triangle edges as springs.
Springs pull points to reach equilibrium.
Mesh Generation & Optimisation
Vertex Shifting – Steepest Gradient
Calculate gradient of TM for each vertex
Step size: smaller than radius of smallest triangle
in-circle.
Move.
If TM did not improve, reduce step size and try
again.
Mesh Generation & Optimisation
Quad Flipping & Vertex Shifting
Animation: https://youtu.be/K6HIjYyjVYU
Results
Matterhorn – 98 Triangles
Animation: https://youtu.be/TqQO4s_Uk2A
Results
Mount St Helens – 98 Triangles
Animation: https://youtu.be/_rBfJfUtr9w
Results
Torrener Joch – 196 Triangles
Animation: https://youtu.be/yvmny5qFfYo
Unfolding
Model Building
References
[1] Michael Garland and Paul Heckbert. Fast Polygonal Approximation of Terrains and Height Fields.
Technical Report CMU-CS-95-181
[2] Alper Üngör. "Off-centers: A new type of Steiner points for computing size-optimal quality-guaranteed
Delaunay triangulations." Computational Geometry 42.2 (2009): 109-118.
[3] Andrey N. Chernikov and Nikos P. Chrisochoides. "Generalized Delaunay mesh refinement: From
scalar to parallel." Proceedings of the 15th International Meshing Roundtable. Springer Berlin Heidelberg,
2006.
[4] Diachin et al., A Comparison Of Two Optimization Methods For Mesh Quality Improvement, Springer
London, 2006
[5] Anna M. Leonowicz and Bernhard Jenny, and Lorenz Hurni. Automated Reduction of Visual
Complexity in Small-Scale Relief Shading, Cartographica (volume 45, issue 1)