Computer Modelling of Fallen Snow by Paul Fearing

1
Computer Modelling of Fallen Snowby Paul
Fearing

• Presented by Luv Kohli
• COMP238
• October 29, 2002

2
The Problem

• Want to generate realistic snowy worlds
• Need to determine
• how much snow falls on a scene
• where this snow accumulates

3
Previous work

• Premoze, et al.
• Mostly concerned with far away landscapes
• Uses digital elevation model enhanced with aerial
  photo
• Much larger scale

4
Basic Algorithm (Fearing)

• Two stages
• Snow accumulation stage
• How much snow accumulates per surface
• Snow stability stage
• Resolves unstable snow surfaces

5
Snow accumulation stage

• Attempt to simulate flake flutter
• Shoot snow particles from launch sites towards
  sky
• Like ray tracing, but not straight lines
• Piecewise linear path towards sky

6
Flake flutter
Circles of varying radius
Random points on circles
Flake path
7
Snow accumulation stage

• 10-15 flakes from each launch site shot upwards
• until blocked hit
• or until sky reached miss
• Gives an idea of launch sites occlusion from sky
• Snow accumulation is used to add snow as 3D
  surfaces above model

8
Importance ordering

• Each launch site given a priority based on
  several factors
• Completeness
• Area
• Neighborhoods
• etc.

9
Occlusion boundaries
10
Launch site meshing

• Sites represented as triangles generated from
  original base scene models
• Initially at least one site per upward-facing
  triangle
• Sites can be merged or refined

11
Sky buckets

• Snowfall should be fairly equal across sky
• Sky divided into grid of equal-area buckets
• Flakes representative area spread across one or
  more buckets on a miss

12
Sky buckets
13
Flake dusting

• Thin dusting of snow difficult to represent as
  3D objects
• Semi-transparent procedural noise textured
  polygons used instead
• Dusting polygons placed slightly in front of 3D
  surface

14
Flake dusting
15
Snow stability stage

• Redistribute accumulated snow into stable
  configuration
• Use angle of repose
• Measure of static friction of a pile of granular
  material

16
Angle of repose (AOR)

• 90º for fresh snow
• 15º for slush
• Can model probability of stability around AOR

17
Stability test

• Compute angle between snow surface on site s and
  neighbors ni lower than s
• If angle too steep to support snow, perform
  obstacle test between s and ni
• Shift snow from s to ni if not blocked
• Repeat until no unstable neighbors or s is empty

18
Obstacle test

• avalanche blocked by scene object
• avalanche blocked by snow on object
• avalanche partially blocked by snow on object

19
Stability termination

• Simulation runs out of time
• All launch sites are stable
• Only a small amount of snow moved during last
  pass
• Most unstable snow resolved during first few
  passes

20
Rain, flour, wind

• Rain can be simulated by setting AOR 0º and not
  allowing any flake flutter
• Feasible for other materials, like flour
• Framework in place for basic wind effects

21
Rain
22
Recent work

• Modeling the Accumulation of Wind-Driven Snow
  Bryan E. Feldman and James F. OBrien

23
(No Transcript)
24
References

• Fearing, P. 2000. Computer modelling of fallen
  snow. In Proceedings of ACM SIGGRAPH 2000, 37-46.
• Feldman, B. E., OBrien, J. F. 2002. Modeling the
  accumulation of wind-driven snow. ACM SIGGRAPH
  2002 Technical Sketch.