SLIDE 1
A Partitioned Approach for Efficient Graph-Based Place Recognition - - PowerPoint PPT Presentation
A Partitioned Approach for Efficient Graph-Based Place Recognition - - PowerPoint PPT Presentation
A Partitioned Approach for Efficient Graph-Based Place Recognition Mattia G. Gollub, Renaud Dub, Hannes Sommer, Igor Gilitschenski, Roland Siegwart Problem Processing 3D point clouds can be computationally expensive. Problem
SLIDE 2
SLIDE 3
Problem Processing 3D point clouds can be computationally expensive. Idea Recognize places on the basis of segment matching.
SLIDE 4
Why segments? Good compromise between local and global descriptors.
SLIDE 5
Why segments? Good compromise between local and global descriptors. Do not rely on the “presence of objects” in the scene.
SLIDE 6
Why segments? Good compromise between local and global descriptors. Do not rely on the “presence of objects” in the scene. Do not rely on a “perfect segmentation”.
SLIDE 7
Why segments? Good compromise between local and global descriptors. Do not rely on the “presence of objects” in the scene. Do not rely on a “perfect segmentation”. Allow for descriptive and compact map representation.
SLIDE 8
SLIDE 9
Ground removal + Euclidean segmentation.
SLIDE 10
Eigen value based features [1].
[1] Weinmann, M., Jutzi, B., & Mallet, C. (2014). Semantic 3D scene interpretation: a framework combining optimal neighborhood size selection with relevant features. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2(3), 181.
SLIDE 11
(1) k-NN retrieval. (2) Random forest classifier trained on separate data.
SLIDE 12
Simple descriptors High fraction of false correspondences. Geometric consistency grouping method [2].
[2] Chen, H., & Bhanu, B. (2007). 3D free-form object recognition in range images using local surface patches. Pattern Recognition Letters, 28(10), 1252-1262.
SLIDE 13
Target map
SLIDE 14
SLIDE 15
Geometric consistency grouping [2]
SLIDE 16
Geometric consistency grouping [2]
Find the largest group of pairwise geometrically consistent correspondences.
[2] Chen, H., & Bhanu, B. (2007). 3D free-form object recognition in range images using local surface patches. Pattern Recognition Letters, 28(10), 1252-1262.
SLIDE 17
Geometric consistency grouping [2]
Find the largest group of pairwise geometrically consistent correspondences.
Method:
1. For each correspondence, initialize a new group.
[2] Chen, H., & Bhanu, B. (2007). 3D free-form object recognition in range images using local surface patches. Pattern Recognition Letters, 28(10), 1252-1262.
SLIDE 18
Geometric consistency grouping [2]
Find the largest group of pairwise geometrically consistent correspondences.
Method:
1. For each correspondence, initialize a new group. 2. For each group, iterate over all the other correspondences. Add the correspondence to the group if it is consistent with all the elements in the group.
[2] Chen, H., & Bhanu, B. (2007). 3D free-form object recognition in range images using local surface patches. Pattern Recognition Letters, 28(10), 1252-1262.
SLIDE 19
Geometric consistency grouping [2]
Find the largest group of pairwise geometrically consistent correspondences.
Method:
1. For each correspondence, initialize a new group. 2. For each group, iterate over all the other correspondences. Add the correspondence to the group if it is consistent with all the elements in the group. 3. Select the biggest group and obtain the localization transformation with RANSAC.
[2] Chen, H., & Bhanu, B. (2007). 3D free-form object recognition in range images using local surface patches. Pattern Recognition Letters, 28(10), 1252-1262.
SLIDE 20
Geometric consistency grouping [2] Worst case asymptotic complexity Can find a suboptimal solution depending on vertices ordering.
SLIDE 21
Graph-based recognition
- Problem represented as a consistency graph:
- Correspondences Vertices
- Consistencies Edges
SLIDE 22
Graph-based recognition
- Problem represented as a consistency graph:
- Correspondences Vertices
- Consistencies Edges
- Solved by maximum clique detection.
SLIDE 23
Graph-based recognition
- Problem represented as a consistency graph:
- Correspondences Vertices
- Consistencies Edges
- Solved by maximum clique detection.
- Identify transformation by least squares (Umeyama method).
SLIDE 24
Graph-based recognition
- Problem represented as a consistency graph:
- Correspondences Vertices
- Consistencies Edges
- Solved by maximum clique detection.
- Identify transformation by least squares (Umeyama method).
Naïve graph construction
SLIDE 25
Graph-based recognition
- Problem represented as a consistency graph:
- Correspondences Vertices
- Consistencies Edges
- Solved by maximum clique detection.
Generally NP-complete
- Identify transformation by least squares (Umeyama method).
Naïve graph construction
SLIDE 26
Partition-based graph construction
SLIDE 27
Partition-based graph construction Observation: Two consistent correspondences must follow
SLIDE 28
Partition-based graph construction Observation: Two consistent correspondences must follow
Target map Local map
SLIDE 29
Partition-based graph construction
SLIDE 30
Maximum clique detection We take advantage of the sparseness of the graph.
[3] Eppstein, D., Löffler, M., & Strash, D. (2010, December). Listing all maximal cliques in sparse graphs in near-optimal time. In International Symposium on Algorithms and Computation (pp. 403-414). Springer, Berlin, Heidelberg.
SLIDE 31
Maximum clique detection We take advantage of the sparseness of the graph. Search for maximum clique as proposed by Eppstein et al. [3].
[3] Eppstein, D., Löffler, M., & Strash, D. (2010, December). Listing all maximal cliques in sparse graphs in near-optimal time. In International Symposium on Algorithms and Computation (pp. 403-414). Springer, Berlin, Heidelberg.
SLIDE 32
Results
SLIDE 33
Results
SLIDE 34
SLIDE 35