Matching Distance and Geometric Distribution Aided Learning Multiview Point Cloud Registration
Not on the map. Why?
Abstract (arXiv)
Multiview point cloud registration plays a crucial role in robotics, automation, and computer vision fields. This paper concentrates on pose graph construction and motion synchronization within multiview registration. Previous methods for pose graph construction often pruned fully connected graphs or constructed sparse graph using global feature aggregated from local descriptors, which may not consistently yield reliable results. To identify dependable pairs for pose graph construction, we design a network model that extracts information from the matching distance between point cloud pairs. For motion synchronization, we propose another neural network model to calculate the absolute pose in a data-driven manner, rather than optimizing inaccurate handcrafted loss functions. Our model takes into account geometric distribution information and employs a modified attention mechanism to facilitate flexible and reliable feature interaction. Experimental results on diverse indoor and outdoor datasets confirm the effectiveness and generalizability of our approach. The source code is available at https://github.com/Shi-Qi-Li/MDGD.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Builds on 26 of its 40 references that are in the map. First posted on arXiv 6 May 2025.
Builds on
26 of 40 references.
References · 26
- 2016Layer Normalizationcited by 5,669
- 2014Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translationcited by 4,057
- 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,117
- 2015Fast R-CNNcited by 2,933
- 2018On the Continuity of Rotation Representations in Neural Networkscited by 1,214
- 2017Large-scale Point Cloud Semantic Segmentation with Superpoint Graphscited by 385
- 20163DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructionscited by 336
- 2020PREDATOR: Registration of 3D Point Clouds with Low Overlapcited by 290
- 2022KISS-ICP: In Defense of Point-to-Point ICP -- Simple, Accurate, and Robust Registration If Done the Right Waycited by 132
- 2020Feature-metric Registration: A Fast Semi-supervised Approach for Robust Point Cloud Registration without Correspondencescited by 109
- 20233D Registration with Maximal Cliquescited by 84
- 2020Learning multiview 3D point cloud registrationcited by 78
14 more
- 2021You Only Hypothesize Once: Point Cloud Registration with Rotation-equivariant Descriptorscited by 68
- 2019NeuRoRA: Neural Robust Rotation Averagingcited by 32
- 2023Robust Outlier Rejection for 3D Registration with Variational Bayescited by 31
- 2019Learning Transformation Synchronizationcited by 29
- 2021HARA: A Hierarchical Approach for Robust Rotation Averagingcited by 23
- 2023Robust Multiview Point Cloud Registration with Reliable Pose Graph Initialization and History Reweightingcited by 19
- 2021FINet: Dual Branches Feature Interaction for Partial-to-Partial Point Cloud Registrationcited by 14
- 2021Learning Iterative Robust Transformation Synchronizationcited by 14
- 2021IMFNet: Interpretable Multimodal Fusion for Point Cloud Registrationcited by 12
- 2022RAGO: Recurrent Graph Optimizer For Multiple Rotation Averagingcited by 10
- 2021On Bundle Adjustment for Multiview PointCloud Registrationcited by 7
- 2022GMF: General Multimodal Fusion Framework for Correspondence Outlier Rejectioncited by 7
- 2020Robust Motion Averaging under Maximum Correntropy Criterioncited by 4
- 2023HybridFusion: LiDAR and Vision Cross-Source Point Cloud Fusioncited by 1
Built on by
Citing papers per year: 2025: 1.
1 paper.