Building Vectorization in TerraScan

1
Building Vectorizationin TerraScan

• Arttu Soininen
• Software developer
• Terrasolid

2
(No Transcript)
3
(No Transcript)
4
(No Transcript)
5
(No Transcript)
6
(No Transcript)
7
Development phase

• Evolving tools
• 60 done
• Initial focus on
• high density laser data sets
• manual approach
• accurate vectorization with the help of images
• Future development focus
• low density laser data sets
• automation
• more generalized vectorization

8
What do we see?

• Airborne system
• can capture roof structures
• does not see all the walls
• Laser returns from walls
• poor measurement angle
• mostly from structures with some horizontal
  surfaces (window openings, balconies, ...)

9
What is accurate from laser data?

• Plane equations of large planar surfaces
• large number of measurement points
• Intersection lines
• points from two surfaces

10
What is poor from laser data?

• Small details
• not enough hits
• Position of edges

11
Input data

• Laser points
• planar roof surfaces
• Optional airborne images
• more accurate positioning of edges
• see the structures
• requires TerraPhoto for MicroStation
• Optional 2D building footprint vectors
• alignment of building planes
• positions of walls

12
Preprocessing

• Laser point classification
• ground
• high objects in one class
• Accurate image orientation
• tie points
• aerial triangulation

13
Basic approach

• Find planar surfaces from laser points
• Detect symmetry and adjust plane equations
• Find boundaries of planes
• Align boundary lines
• using images
• to intersection lines
• with each other

14
Monoscopic measurement

• Roof plane equation is known
• Camera orientation is known
• We can measure points on the plane using one raw
  image

15
Resulting vector model

• Simplified vector model ignores
• small forms and details
• structural imperfections
• Roof planes
• top planes
• side planes for thickness
• bottom planes
• roof color from images
• Walls
• start below ground
• position guessed or from footprint

16
Plane direction

• Each plane has a base direction
• direction group has a common base direction
• Base direction affects
• surface equation -- aligns normal vector
• boundary lines -- aligned if rectangle/rectangular
  
• Direction vector is computed
• from normal vector for sloped planes -- accurate
• from point cloud for flat planes -- less accurate

Direction
Section
Top
17
Limitations

• Tools are not usable if roof
• Has no planar surfaces
• Contains only small planes and point density is
  not high enough
• Gives no returns (glass or dark surface)

18
Easy
19
Normal
20
Difficult
21
Almost impossible