LIDAR Accuracy on Asphalt Road

1
LIDAR Accuracy on Asphalt Road

• Arttu Soininen
• Terrasolid Ltd

2
Case story

• Task
• produce an accurate asphalt road surface model
• Purpose
• asphalt is going to be resurfaced
• machinery is going to be automatically guided by
  total station positioning
• Note
• we assume that total station network is perfect
• 481 total station points on asphalt for control
• surrounding terrain is not important

3
Coordinate setup

• TerraScan uses integer coordinate system
• Sentimeter steps OK for general terrain mapping
• Use millimeter or 1/10 of a millimeter for best
  accuracy work

4
Laser scanning

• 15 km of road
• TopEye measured in October 2003 with a scanner
  measuring about 7000 points per second
• Digital camera images with 2 cm resolution
• Flown in two directions at 100 m altitude

5
Initial laser data accuracy

• IMU / scanner misalignment angles carefully
  calibrated from data set
• TerraMatch gives no real HRP improvement
• Average difference between surfaces from
  different flightlines 3.689 cm
• Average difference on asphalt 3.440 cm

6
Dz correction for whole flightlines

• TerraMatch gave dz corrections for whole
  flightlines
• -1.7 cm line 1
• 0.7 cm line 2
• 0.8 cm line 3
• 0.4 cm line 4
• -0.2 cm line 5
• Average difference between surfaces from
  different flightlines after correction 3.299 cm,
  on asphalt 2.997 cm

7
Fluctuating elevation correction

• Corrects for inaccuracy of trajectory elevations
• TerraMatch computed elevation difference of each
  flightline to others at 1 second intervals
• Each 1 second interval was corrected with the
  average of 3 consecutive seconds
• Correction limited to max 2 cm
• Average difference between surfaces from
  different flightlines 3.070 cm

8
Find Fluctuations

• Correction will modify laser points of each
  interval with a unique dz correction
• User can select
• how correction curve is averaged from consecutive
  intervals
• what is the maximum correction to apply

9
Find Fluctuations - Before
10
Find Fluctuations - After
11
Cutting edges of scan lines

• TerraScan cut edges of scan lines where accuracy
  is not as good as at the center of scan lines

12
Geoid correction

• Transform from GRS80 ellipsoid to orthometric
  height
• Average elevation difference between LIDAR
  surface and total station points was computed for
  each 1 km interval

13
Smoothing of laser surface

• Classify points within 10 cm from ground to
  ground
• Smoothen laser surface
• Std dev asphalt against total station points 2.32
  cm

14
Smoothing 10 cm spread

• Improves accuracy on hard surfaces
• Requires interactive work -- draw polygons
• Raises laser data on hard surfaces
• systematic biases on different surfaces closer
• Example before
• laser data 5 cm too high on asphalt
• laser data 9 cm too high in terrain
• Example after
• laser data 7 cm too high on asphalt
• laser data 9 cm too high in terrain

15
Breaklines on asphalt

• 2D breaklines drawn on orthophoto were draped to
  laser point surface
• smoothenes variations in longitudinal direction
• Std dev of resulting TIN model
• fix points 2.02 cm, min -5.60, max 5.30

16
(No Transcript)
17
Summary of steps

• Match laser strips internally
• HRP misalignment ? 3.440 cm
• Dz per flightline ? 2.997 cm
• Elevation fluctuations ? 2.705 cm
• Cut overlap
• Classify ground
• Geoid correction based on local points ? 2.48 cm
• Classify 10 cm spread and smoothen ? 2.32 cm
• Drape breaklines on laser surface
• Model breaklines surrounding terrain ? 2.02 cm