Structuring, Learning, and Innovating Multi-Nationally Dr. Ellen A. Drost.
Detecting Geometric Primitives in 3D...
Transcript of Detecting Geometric Primitives in 3D...
![Page 1: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/1.jpg)
© 2003-2018 MVTec Software GmbH | Any use of content and images outside of this presentation or their extraction is not allowed without prior permission by MVTec Software GmbH
Detecting Geometric Primitives in 3D Data
![Page 2: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/2.jpg)
Background: Building Blocks for Point-Pair-Voting
Feature 3D Point Pairs
Parameter Space Local, Data-Driven Parameters
Feature Matching Implicit Model Description
Detection Voting Scheme
![Page 3: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/3.jpg)
Background: Point Pair Features
Fast, invariant, discriminative,
define local reference frame
![Page 4: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/4.jpg)
Background: Point Pair Feature Database
Find similar point pairs in O(1)
![Page 5: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/5.jpg)
Background: Local Pose Parameters
Rigid 6D pose:
Large, complex
parameter space
![Page 6: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/6.jpg)
Background: Local Pose Parameters
Rigid 6D pose:
Large, complex
parameter space
Assume one 3D point to
be aligned
![Page 7: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/7.jpg)
Background: Local Pose Parameters
Rigid 6D pose:
Large, complex
parameter space
Assume one 3D point to
be aligned
Fix scene point
(“reference point”)
Find corresponding
model point and rotation
around normal vector
![Page 8: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/8.jpg)
Background: Local Pose Parameters
![Page 9: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/9.jpg)
For each scene point, find best local parameters (corresponding model point, rotation angle)
through voting
Initialize accumulator array with zeros
Iterate over other points
For each point pairs, vote
Voting Scheme
![Page 10: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/10.jpg)
Select a 3D scene point, create the local voting space
Voting Scheme
![Page 11: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/11.jpg)
Pair the 3D point with all other 3D points
Voting Scheme
![Page 12: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/12.jpg)
Compute the point pair feature
Voting Scheme
![Page 13: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/13.jpg)
Find corresponding point pairs on the model
Voting Scheme
![Page 14: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/14.jpg)
Vote for every possible correspondence
Voting Scheme
![Page 15: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/15.jpg)
Maximum in the voting space is the locally optimal pose
Voting Scheme
![Page 16: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/16.jpg)
Geometric primitives often arise in practical applications
Calibration using spheres
Find background planes
Navigation
Remove background plane before object detection
Raw cylinders with varying radii (rigid shape is too constraining)
The base method works for primitives, but local parameters contain redundancies
that make the voting slower
Propose to adapt the method to
Remove redundancies
Exploit explicit nature of primitives for faster feature matching
Add shape parameters (radius / scale)
Geometric Primitives
![Page 17: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/17.jpg)
Optimizations for Geometric Primitives
Feature 3D Point Pairs
Parameter Space Local, Data-Driven Parameters
Feature Matching Implicit Model Description
Detection Voting Scheme
![Page 18: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/18.jpg)
Geometric Primitives: Using Symmetry Information
![Page 19: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/19.jpg)
Implicit Point Pairs for Planes
![Page 20: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/20.jpg)
Implicit Point Pairs for Spheres
![Page 21: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/21.jpg)
Implicit Point Pairs for Cylinders
![Page 22: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/22.jpg)
Pipeline
![Page 23: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/23.jpg)
Results on the SegComp ABW Dataset
![Page 24: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/24.jpg)
Results on a Synthetic Dataset
![Page 25: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/25.jpg)
Results on Synthetic Data
![Page 26: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/26.jpg)
Results on Real Data
![Page 27: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/27.jpg)
Results on Real Data
![Page 28: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/28.jpg)
Outlook: Solids of Revolution
![Page 29: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/29.jpg)
Iterative Voting though Graph Matching
Feature 3D Point Pairs
Parameter Space Local, Data-Driven Parameters
Feature Matching Implicit Model Description
Detection Voting Scheme
![Page 30: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/30.jpg)
Graph Matching
correspondences between
model and scene points
Correct correspondences form a dense subgraph
Duchenne, O., Bach, F., Kweon, I.S., Ponce, J.: A tensor-based algorithm for high-order graph matching. PAMI
33(12) (2011)
connect
consistent correspondences
![Page 31: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/31.jpg)
Graph Matching
correspondences between
model and scene points
Correct correspondences form a dense subgraph
connect
consistent correspondences
Assignment vector:
Relax:
![Page 32: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/32.jpg)
Graph Matching
correspondences between
model and scene points
Correct correspondences form a dense subgraph
connect
consistent correspondences
Solved using gradient descend:
Equivalent to the power iteration with proven convergence
Equivalent to multiple rounds of voting
Assignment vector:
Relax:
![Page 33: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/33.jpg)
Graph Intuition
![Page 34: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/34.jpg)
Graph Pruning
![Page 35: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/35.jpg)
Results
![Page 36: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/36.jpg)
Results
![Page 37: Detecting Geometric Primitives in 3D Datacmp.felk.cvut.cz/sixd/workshop_2018/data/drost_r6d_eccv...Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition;](https://reader033.fdocuments.net/reader033/viewer/2022041912/5e67db7ade517600cf1bd5f5/html5/thumbnails/37.jpg)
Bertram Drost, Slobodan Ilic: Local Hough Transform for 3D Primitive Detection; in:
International Conference on 3D Vision (3DV), 398-406, 2015.
Bertram Drost: Point Cloud Computing for Rigid and Deformable 3D Object Recognition; PhD
Thesis, Faculty of Informatics, Technical University of Munich, 2016.
Bertram Drost, Slobodan Ilic: Graph-based deformable 3d object matching; in: German
Conference on Pattern Recognition, 2015.
Adam Hoover, Gillian Jean-Baptiste, Xiaoyi Jiang, Patrick J. Flynn, Horst Bunke, Dmitry B.
Goldgof, Kevin W. Bowyer, David W. Eggert, Andrew W. Fitzgibbon, Robert B. Fisher: An
Experimental Comparison of Range Image Segmentation Algorithms; in: IEEE Transactions on
Pattern Analysis and Machine Intelligence, 18 (7):673–689, 1996.
References