Reconstructing 3D Tree Models from Instrumented Photographs

1
Reconstructing 3D Tree Models from Instrumented
Photographs

• Ilya Shlyakhter ,Max Rozenoer,Julie Dorsey, and
  Seth Teller
• Massachusetts Institute of Technology
• IEEE Computer Graphics and Applications
• 2001

2
Realism

• The realistic modeling of vegetation is an
  important problem in computer graphics.
• Vegetation adds a significant dimension of
  realism to a scene.
• Improve realism in architectural landscapes by
  trees.

3

• Many techiques and modeling package have been
  developed for contructing a tree of a particular
  type (Tree Professional Software, MaxTrees etc.).
• Variation between two tree of single type can be
  very significant, depending on factors (growth
  conditions - sunlight, human intervention -
  pruning branch)

4

• The techniques and packages currently available
  dont provide sufficient control over the final
  shape.
• The author proposed a solution for reconstructing
  a faithful 3D model of foliaged tree from a set
  of instrumented photographs.

5
System Details

• The system consists of four stages.
• The input is a set of images of a tree, 4-15
  images covering at least 135 degree around the
  tree
• Assume the position and orientation for each
  image is known.

6
Stage

• 1 Image segmentation
• 2 Visual hull construction
• 3 Skeleton construction
• 4 L-system

7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
(No Transcript)
14
Future work

• Involves tree-specific knowledge into the
  algorithm.
• Recover foliage density in the tree.
• The resulting skeleton often differs greatly from
  the actual one. Adding constraints based on
  knowledge of growth patterns for particular tree
  types should mitigate this problem.

15
Future work (cont.)

• The L-system used for reconstruction should be
  tree-type specific.
• No work has been done on reconstructing the
  nonfoliaged winter trees.