The fusion process of LiDAR and map data to generate 3D city and landscape models

1
The fusion process of LiDAR and map data to
generate 3D city and landscape models

• Sander oude elberink
• Geospatial world forum 16 May 2013

2

• Generation of nationwide 3D city and landscape
  models using national datasets
• 11.000, BGT, fused with AHN-2 (8 p/m2)
• 110.000, TOP10, fused with AHN-2 (8 p/m2)
• Fusion process
• Research questions
• Hydrocity (11.000)
• 3DIMGeo (11.000)
• 3DTOP10NL (110.000)

3

• Example in figures

4

• buildings

5

• The questions

• Which lidar points have to be used to transfer
  the height to an object?
• How to use the semantics of the map data?
• How to assign a height to a point, boundary or
  surface?
• What is the quality of that height?
• How to deal with noise in both the map and lidar
  data?

6

• Fusing map and laser data

7

• Select laser points per map point, per polygon

• Transfer height from selected points to map point
• In general resulting in at least 2 heights per
  map point.
• What to do with the differences?
• Semantics between classes

8

• Hydrocity

• Produce 3d model for hydrological applications

• See also presentation of Mark Kroon Neo
• Aim was to keep small relative height differences
• (but not the ones caused by noise)
• Curbstones
• Boundary between 2 infrastructural polygons
  (road, sidewalk).
• Function of object in addition to class label

9

• Hydrocity

• Object based

• Per object height, infiltration capacity,
  surface roughness
• Interpolated to grid for run off modelling

10

• 3DIMgeo

• 11.000

• As a product of 3D Pilot, start of 3D SIG NL (see
  presentation of Jantien Stoter).
• Based on IMGeo, CityGML standards.
• Workbench in FME, in cooperation with con terra
  GmbH (Christian Dahmen).
• LoD0, LoD1 and LoD2.

11

• FME - 3DIMGEO tools - FME

• Use FME to (summary)
• Read and validate source data CityGML 2D IMGeo
  LiDAR (AHN-2)
• 'Point-On-Polygon' operation (assign laser data
  to polygons)
• Run manage the complete workflow -gt Single User
  Interface
• Use '3D IMGeo tools' developed by U Twente to
• Prepare map data and laser data for the 3D
  reconstruction.
• Assign height to the map boundaries for a LoD0
  terrain description.
• Assign a height description inside the 3D
  polygons. Results are TIN surfaces at LoD0.
• Calculate LoD1 or LoD2 buildings and forest
• Use FME again to write result data CityGML 3D
  IMGeo

12
(No Transcript)
13

• 3Dtop10nl

• 110.000
• Fused with AHN-2 (8 p/m2)

14

• Implications of fusion

• TOP10NL topographic representation, geometric
  accuracy 2 m
• AHN-2 geometric 3D representation, geom acc lt
  0.5 m, 8-10 p/m2
• Aim for selecting correct points
• Do we need all laser points?

15
(No Transcript)
16
(No Transcript)
17

• Rules to calculate object height

Class Lidar data taken from 3D Representation type / Semantic constraint Initial height of object points on boundary Surface description
Water Ground Horizontal plane All object points are set to average height Determined by triangulation of boundary object points
Roads Ground Locally planar Each object point is determined by height of local fitted plane Determined by triangulation of boundary object points
Terrain Ground May vary locally Each object point is determined by height of local fitted plane Lidar points are inserted inside polygon, followed by constrained triangulation
Buildings Non-ground Horizontal plane, LoD 1 All object points are set to average height Determined by triangulation of boundary points
Forest Non-ground May vary locally Each object point is determined by height of local fitted plane Lidar points are inserted inside polygon, followed by constrained triangulation
18

• Rules to combine height of neighbouring polygons

  Water Road Terrain Building Forest
Water Both keep own height Both own height, create additional polygon below road Take water height Both keep own height, create wall in-between Both keep own height, create wall in-between
Road   Average if close in height Take road height Both keep own height, create wall in-between Both keep own height, create wall in-between
Terrain     Take average of both heights Both keep own height, create wall in-between Both keep own height, create wall in-between
Building       Both keep own height Both keep own height
Forest         Both keep own height
19
(No Transcript)
20

• The questions

• And the frustrating answers

• Which lidar points have to be used to transfer
  the height to an object?
• Depends on the object.
• How to use the semantics of the map data?
• Depends on the map/application.
• How to assign a height to a point, boundary or
  surface?
• Depends on the object.
• What is the quality of that height?
• Depends on the workflow.
• How to deal with noise in both the map and lidar
  data?
• Deal with it.

21

• Near future

• Kadaster will go for 3DTOP10NL
• 3D IMGeo tools are open (since April 2013) and
  integrated into FME
• Nice link between Geo practice and research

22

• More info