This tutorial describes the processing required to classify a 3D point cloud (LiDAR) into the following classes: undefined, ground, Vegetation (low,mid,high), building, powerline. It uses the LIS Software from Laserdata Company (www.laserdata.at).
2. 2
The Data Set
The data set is delivered as compressed ASPRS LAS file (*.laz) with basic attributes like
gps-time, return number and intensity. This tutorial describes the processing required to
classify the points into the following classes:
- undefined
- ground
- vegetation (low/mid/high)
- building
- power line
3. 3
Import and Visualization of the Point Cloud
First of all, we import the data set with the following tool:
Tools > LASERDATA LiS > Import/Export > Import LAS/LAZ Files
Provide the
data_heli.laz file as
input.
Then check those
attributes you like
to import.
4. 4
Import and Visualize Point Cloud
A double left-click on the dataset in the Data tab opens up a Map View:
5. 5
3D Visualization
Let's have a quick look at the data set in 3D. We can use the Point Cloud Viewer for this
purpose:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
Provide the point
cloud as input.
Also provide the
“intensity” attribute
with the Intensity
parameter.
6. 6
3D Visualization
Press shortcut key “2” to change from elevation to intensity colored (use shortcut key “1”
to switch back to elevation). We can use this shortcut because we specified the intensity
attribute in the parameter settings. Finally quit the Point Cloud Viewer by closing its
window.
7. 7
Ground Classification
The first step in point cloud classification is to separate the points into ground and non-
ground points.
Tools > LASERDATA LiS > Classification > Ground Classification (PC)
Provide the point
cloud as input.
Some parts of the
terrain are quite
steep, so we adjust
the Maximum Angle
to 30°
8. 8
3D Visualization
Let's have a look at the ground classification result in 3D:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
The tool remembers the
last setting, so we only
have to provide the new
“grdclass” attribute with the
Classification parameter.
9. 9
3D Visualization
Press shortcut key “5” to change from elevation to classification colored (use shortcut key
“1” to switch back to elevation; use shortcut key “6” to show only ground points). We can
use this shortcut because we specified the classification attribute in the parameter
settings. Finally quit the Point Cloud Viewer by closing its window.
10. 10
Point Cloud Segmentation
Next, as a prerequisite to building classification, we do a segmentation of the non-ground
points into planar objects.
Tools > LASERDATA LiS > Segmentation > Segmentation by Plane Growing (PC)
Provide the point cloud
as input. We use the
ground classification as
threshold attribute (so
we only segmentize
non-ground points)
We need to adjust some
parameters: we can use
stricter settings as this
is a helicopter data set
with a higher point
density
We only process point class 1 (undefined)
11. 11
3D Visualization
Let's have a look at the segmentation result in 3D:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
Provide the new “segmentid”
attribute with the Random
parameter. This allows us to
colorize the segments by
random.
12. 12
3D Visualization
Press shortcut key “q” to change from elevation to random colored (press “q” again to
assign another set of random colors). Values of -1 indicate that the point was skipped
during segmentation, points skipped in region growing have a segment ID of 0. The
individual segments group points on planar surfaces.
13. 13
Building and Vegetation Classification
Next, we classify building and vegetation points.
Tools > LASERDATA LiS > Classification > Enhanced Point Cloud Classification
Provide the point cloud
as input. We build upon
already derived
attributes: segmentid,
ground classification,
and the height above
ground
Again we need to adjust
some parameters due
to the higher point
density
14. 14
3D Visualization
Let's have a look at the building and vegetation classification result in 3D:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
Change the attribute provided
with the Classification
parameter from “grdclass” to
“classid”.
15. 15
3D Visualization
Press shortcut key “5” to change from elevation to classification colored; use shortcut key
“1” to switch back to elevation; use shortcut keys “6, 7, 8, 9, 0” to change to further
visualization options.
16. 16
Building Refinement
As you may have noticed, there are some points on building roofs still classified as
vegetation. We will clean this up using a majority filter, which is constrained to work only
on building and vegetation points.
Tools > LASERDATA LiS > Filtering > Majority Filter (PC)
Provide the point cloud
as input and select
“classid” as attribute to
filter.
Adjust the radius,
constrain the filtering to
high vegetation points
(classid 5), and exclude
undefined, ground and
lower vegetation points
from the computation.
17. 17
3D Visualization
Let's have a look at the refined building classification in 3D:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
Change the attribute provided
with the Classification
parameter from “classid” to
“majority_classid”.
18. 18
3D Visualization
Press shortcut key “5” to change from elevation to classification colored. You can change
to the raw classification by changing the Color Attribute from “majority_classid” to
“classid”.
19. 19
Power Line Classification
Next step is a preliminary power line classification.
Tools > LASERDATA LiS > Classification > Power Line Classification
Provide the point cloud as input
and select “majority_classid”
with the Classification
parameter and “dz” with the
Threshold Attribute parameter.
We will use the latter to
constrain the classification to
points higher than 2.5 m.
Also adjust the Search Radius
parameter used for line
detection to 1.5 m.
20. 20
3D Visualization
Let's have a look at the preliminary power line classification result in 3D:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
Change the attribute provided
with the Classification
parameter from
“majority_classid” to
“classid_pl”.
22. 22
Power Line Refinement
As you may have noticed, some points in vegetation and on building facades are detected
as power line. We will clean this up using a refinement filter, which also adds a cable ID to
each power line. The latter can be used for catenary curve fitting for example.
Tools > LASERDATA LiS > Classification > Power Line Refinement
Provide the point cloud
as input and select the
“classid_pl” and “dz”
attributes.
Adjust the search radii
and the line tolerance.
23. 23
3D Visualization
Let's have a look at the refined power line classification in 3D:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
Change the Classification parameter to
“classid_pl_refined” and the Random
parameter to “cableid_pl_refined”.
25. 25
3D Visualization
Press shortcut key “q” to change to random colored, showing the cable ID (press “q” again
to assign another set of random colors).
26. 26
Final Classification
As as last step, we will clean up the building facades by labeling all vegetation points
(classid 3 to 5) to undefined (classid 1).
Tools > LASERDATA LiS > Classification > Clean Building Facades (PC)
Provide the point cloud
as input and select the
“classid_pl_refined”
Adjust the search radius
and set the Target Class
to 1.
27. 27
3D Visualization
Let's have a look at the final classification result in 3D:
Tools > LASERDATA LiS > Visualization > Point Cloud Viewer [interactive]
Change the Classification parameter to
“classid_pl_refined_final”