Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

34
Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani

Transcript of Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Page 1: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Automated Reconstruction of Industrial Sites

Frank van den HeuvelTahir Rabbani

Page 2: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Overview

• Introduction• Automation: how does it work?• Sample project off-shore platform• Accuracy• Future• Conclusions

Page 3: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

The groupPhotogrammetry & Remote Sensing• “Development of efficient techniques for the

acquisition of 3D information by computer-assisted analysis of image and range data“

Page 4: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

The projectServices and Training through Augmented Reality (STAR)

• EU fifth framework – IST programme• “Develop new Augmented Reality techniques

for training, on-line documentation, maintenance and planning purposes in industrial applications”

• AR-example: virtual human in video

Page 5: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

The projectServices and Training through Augmented Reality (STAR)

• Partners: Siemens, KULeuven, EPFL, UNIGE, Realviz

• TUDelft: “Automated 3D reconstruction of industrial installations from laser and image data”

Page 6: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Automated reconstruction procedureOverview (1/3)• Segmentation• Grouping points of surface patches

Page 7: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Automated reconstruction procedureOverview (2/3)• Segmentation• Grouping points of surface patches

• Object Detection• Finding planes and cylinders

Page 8: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Automated reconstruction procedureOverview (3/3)• Segmentation• Grouping points of surface patches

• Object Detection• Finding planes and cylinders

• Fitting• Final parameter estimation

Page 9: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Segmentation – step 1

• Estimation of surface normals using K-nearest neighbours (here K=10 points)

Page 10: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Segmentation – step 2

• Region growing using:• Connectivity (K-nearest

neighbours) • Surface smoothness

(angle between normals)

Page 11: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Detection – Planes

• Plane detection using Hough transform• Find orientation as maximum on Gaussian

sphere

Page 12: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Detection – Cylinders

• Cylinder detection using Hough transform in 2 steps:• Step 1: Orientation

Page 13: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Detection – Cylinders

• Cylinder detection using Hough transform in 2 steps:• Step 1: Orientation

Page 14: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Detection – Cylinders

• Cylinder detection using Hough transform in 2 steps:• Step 1: Orientation (2 parameters)• Step 2: Position and Radius (3 parameters)

u,v search space at correct Radius

Page 15: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Example: detection of two cylinders

• Point cloud segment

Page 16: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Example: detection of two cylinders

• Surface normals

Page 17: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Example: detection of two cylinders

• Normals on Gaussian sphere

Page 18: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Example: detection of two cylinders

• Orientation of first cylinder (next: position)

Page 19: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Example: detection of two cylinders

• Remove first cylinder points from segment

Page 20: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Example: detection of two cylinders

• Procedure repeated for second cylinder

Page 21: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Example: detection of two cylinders

• Result: two detected cylinders

Page 22: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Fitting

• Complete CSG model + constraint specification

• Final least-squares parameter estimation of CSG model

Page 23: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Fitting

• Final least-squares parameter estimation of CSG model• Minimise sum of squared distances• Enforce constraints

Page 24: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Results on platform modelling

• Scanned by Delftech in 2003• Subset of 17.7 million points used by TUD:• Automated detection of 2338 objects• R.M.S. of residuals 4.3 mm

Page 25: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Results on platform modelling

Page 26: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Results on platform modellingStatistics

• Points: 17.7 million• Points in segments: 14.2 million(80%)• Points on objects:9.3 million (53%)• Detected:• Planar patches: 946• Cylinders: 1392

• Data reduction:• Object parameters 9798• 500 Mb to 0.1 Mb

Page 27: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

0 2 4 6 8 10 12 14 16 18 200

10

20

30

40

50

60

70

80

90

100

Residual (mm)

%ag

e of

poi

nts

Cumulative histogram of residuals

Results on platform modelling Accuracy• Residual analysis:

• RMS: 4.3 mm• 83% < 5 mm• 96% < 10 mm

Page 28: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Accuracy

• Data precision:• Scanner: 6 mm (averaging: 3 mm)

• Scanner dependent

• Model precision:• Discrepancies models - real world: 0.1-10 mm ?

• Limited production accuracy• Deformations• Imperfections in segmentation

Page 29: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Accuracy

• Object deformation or segmentation limitations?

Fitting after initial segmentation

Max.residual 21 mm

Fitting after rejecting large residuals

Max. residual 9 mm

Page 30: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Future – automation

• Reconstruction using laser data:• Segmentation, primitive detection (available)• Correspondence between primitives >

registration• Model improvement:

• Constraint detection (piping structure)• Recognition of complex elements in a database

• Integration with digital imagery

Page 31: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Future – integration with imagery

• Instrumentation developments• Scanners with integrated high-resolution digital

camera

• Accuracy improvement• Imagery complementary: Laser for surfaces, image for

edges• Integrated fitting of models to laser and image data

Page 32: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Future – integration with imagery

• Instrumentation developments• Scanners with integrated high-resolution camera

• Accuracy improvement• Imagery complementary: Laser for surfaces, image for

edges• Integrated fitting of models to laser and image data

• Flexibility of image acquisition: Completeness• Non-geometric information (What is there?)

Page 33: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Future – integration with imagery

Page 34: Automated Reconstruction of Industrial Sites Frank van den Heuvel Tahir Rabbani.

Conclusions

• Bright future for automation using laser data• More research to be done:• Automated registration• Integration with digital imagery• Using domain knowledge for automated

modelling:• Closer connection to the model users needed:• Domain knowledge for automation• Utilisation of research results