Derived Point Clouds - Knowledge...

9
21-Jun-17 1 Photogrammetric Point Cloud Processing Jim Maloney GeoCue Group, Inc. 9668 Madison Blvd., Suite 202 Madison, AL 35758 +1 (256) 461-8289 [email protected] support.geocue.com GeoCue Terrasolid LIDAR Training Terrasolid Training Webinar – 22 Jun 2017 Derived Point Clouds Noisy Data Very Dense- 200-500 pts/m 2 . Lack of points below Vegetation 05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 2

Transcript of Derived Point Clouds - Knowledge...

21-Jun-17

1

Photogrammetric Point Cloud Processing

Jim Maloney

GeoCue Group, Inc.

9668 Madison Blvd., Suite 202

Madison, AL 35758

+1 (256) 461-8289

[email protected]

support.geocue.com

GeoCue Terrasolid LIDAR TrainingTerrasolid Training Webinar – 22 Jun 2017

Derived Point Clouds

• Noisy Data

• Very Dense- 200-500 pts/m2.

• Lack of points below Vegetation

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 2

21-Jun-17

2

Pre-Processing Macro

• Echo information and surface roughness are not as useful with ground

classification with photogrammetric point clouds, we need to take a

different approach to deal with the noise than we do with normal

LIDAR data

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 3

Classify Isolated Points

• Isolated points within a photogrammetric point cloud are never useable

and almost always not meaningful.

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 4

21-Jun-17

3

Classify Surface

• Classifies suitable points into a Potential surface class for further

examination by identifying the median points within this noisy surface.

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 5

Smoothen XYZ

• Modifies the XYZ positions of the Model Keypoints to produce a more homogenous representation of surfaces, both horizontal and vertical.

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 6

Surface Points

Smoothed Surface Points

21-Jun-17

4

Thin Points

• Reduces the density of the points to make the dataset more

manageable.

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 7

Raw Point Cloud

Processed Point Cloud

Classify Hard Surface

• Classifies tentative ground in a similar manner as the Classify Surface

routine. It classifies points on a plane-of-best-fit within a tolerance

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 8

21-Jun-17

5

Classify Ground

• Classifies the final ground class in preparation for above ground

feature classification processing

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 9

Photogrammetric Point Cloud Workflow• Pre-Processing and data review

• Classify Isolated Points

• Classify Potential Surface (Surface Routine)

• Smoothen Potential Surface

• Thin Potential Surface

• Classify Tentative Surface (Hard Surface Routine)

• Classify Ground (Ground Routine)

17-21 Apr 2017 GeoCue/Terrasolid Training – Mercer 11

21-Jun-17

6

Classification Using Groups

• Run grouping of above ground

points

• Goal is to have each object as one

group of points

• Software stores group value for

each point in FastBinary file format

• Manual and automatic classification

can work on object level

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 12

Group Classification Macro

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 13

21-Jun-17

7

Compute Distance

• Computes the distance between points and the ground.

• This distance information is stored for use in other routines.

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 14

Classify by Distance

• Classifies points into medium and high vegetation based on the

distance values computed in the previous step.

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 15

21-Jun-17

8

Assign Groups

• Builds groups from points in source classes

• Typically high vegetation or medium+high vegetation

• Can use four different grouping principles:

• Group planar surfaces finds large enough planar surfaces such as

roofs or walls

• Group by tree logic finds groups using watershed algorithm starting

from highest local point

• Group by density uses spacing between points

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 16

Classify Groups / By best match

• Software can evaluate each group using multiple object recognition

routines

• Classifies each group to best matching class

Example: software may evaluate one group to be:

• Building roof with 0% probability

• Building wall with 0% probability

• Tree with 77% probability

• Pole with 42% probability

• Vegetation with 58% probability

• Car with 0% probability

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 17

21-Jun-17

9

Processing Steps for Airborne LIDAR

• Classify ground

• Classify wires if needed

• Use compute distance to compute height above ground value for each point

• Classify medium vegetation using 'Classify / By distance'

• Classify high vegetation using 'Classify / By distance'

• Group points using 'Group / Assign groups'

• Classify groups using 'Group / Classify / By best match' and other group based

routines

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 18

Exercise #1800

05-07 Jun 2017 GeoCue/Terrasolid Training – Mercer 19

Complete the following exercise using the training set data (or your own data).

Please refer to handout for full details:

1. Load (or reload) the laser points from your GeoCue Sample Project.

2. Classify Isolated Points

3. Classify Surface

4. Smoothen XYZ

5. Thin Points

6. Classify Hard Surface

7. Classify Ground.

8. Compute distance

9. Classify by Distance

10. Assign Groups

11. Classify by Best Match