pith. sign in

arxiv: 2508.17034 · v2 · pith:3WVSIA73new · submitted 2025-08-23 · 💻 cs.RO · cs.CV

DualReg: Dual-Space Filtering and Reinforcement for Rigid Registration

Jiayi Li , Yuxin Yao , Qiuhang Lu , Juyong Zhang This is my paper

Pith reviewed 2026-05-21 22:35 UTC · model grok-4.3

classification 💻 cs.RO cs.CV
keywords rigid registrationpoint cloud registrationfeature matchingRANSACgeometric proxiesreal-time registrationKITTI dataset
0
0 comments X

The pith

A dual-space method filters feature matches with one-point RANSAC then refines them via geometric proxies to register rigid transformations efficiently.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper sets out to show that rigid registration can exploit both feature-based matching for handling large transformations and local geometry for fine alignment by processing them in separate spaces. It first runs a lightweight one-point RANSAC step plus refinement to discard unreliable feature correspondences, then treats the survivors as anchor points, builds geometric proxies from them, and solves a tailored objective for the transformation. A reader would care because noisy, partially overlapping scans appear in robotics and 3D reconstruction, where pure feature methods lack precision and pure geometry methods need good initialization and run slowly. If the approach works, it delivers real-time speeds on standard benchmarks without losing accuracy.

Core claim

By introducing an efficient filtering mechanism of a computationally lightweight one-point RANSAC algorithm and a subsequent refinement module to eliminate unreliable feature-based correspondences, then treating the filtered correspondences as anchor points to extract geometric proxies and formulate an effective objective function with a tailored solver, the transformation can be estimated accurately and quickly for rigid registration of noisy, partially overlapping data.

What carries the argument

The dual-space paradigm that first applies one-point RANSAC filtering to feature correspondences and then uses the surviving matches as anchors to extract geometric proxies for objective-function optimization.

If this is right

  • Large transformation differences can be handled quickly while still reaching fine local alignment on partially overlapping scans.
  • Real-time rigid registration becomes practical for noisy data without requiring a strong initial guess from another module.
  • The same anchor-point idea supports efficient solvers that keep accuracy comparable to heavier geometry-only baselines.
  • Applications such as robotics mapping gain substantial CPU-time savings while preserving output quality on standard outdoor datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same filtering-plus-proxy pattern could be tested on indoor or LiDAR-camera fusion datasets to check whether the speedup generalizes beyond KITTI.
  • Replacing the hand-crafted geometric proxies with learned local descriptors might tighten the accuracy gap on very sparse overlaps.
  • The anchor-point concept suggests a possible route for extending the method to incremental registration inside SLAM pipelines.

Load-bearing premise

The one-point RANSAC filtering step can reliably remove unreliable feature-based correspondences so the surviving matches serve as effective anchor points for the subsequent geometric-proxy refinement.

What would settle it

Running the method on the KITTI dataset after disabling or replacing the one-point RANSAC filter and observing that registration error rises sharply or the claimed speedup disappears would show the dual-space claim does not hold.

Figures

Figures reproduced from arXiv: 2508.17034 by Jiayi Li, Juyong Zhang, Qiuhang Lu, Yuxin Yao.

Figure 1
Figure 1. Figure 1: Registration Recall and Average Runtime (per-registration) on [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Pipeline of the proposed method. We propose DualReg, a dual-space paradigm for robust rigid registration. We first propose an efficient filtering [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Correspondences meet symmetry. cases. Specifically, we compute the orthogonal matrix M∗ = arg min M∈R3×3 X ci∈I(cj ) ∥Mvi + tM − ui∥ 2 , s.t. MTM = I, (6) where I ∈ R 3×3 is the identity matrix. tM is the translation vector and we don’t need to solve it here. Similar to solving the rotation matrix [49], we can compute Singular Value Decomposition (SVD) UsΣVs T = X ci∈I(cj ) (vi − v)(ui − u) T , where v = X… view at source ↗
Figure 3
Figure 3. Figure 3: Tangential distance constraint. If two correspondences belong to the [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Iterative process of one-point RANSAC. Each circle in the figure () [PITH_FULL_IMAGE:figures/full_fig_p005_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison on 3DMatch. The yellow and blue point clouds represent the source and target point clouds, respectively. The first two rows [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative comparison on 3DLoMatch. The yellow and blue point clouds represent the source and target point clouds, respectively. The first two [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative comparison on KITTI. The yellow and blue point clouds represent the source and target point clouds, respectively. The first two rows [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
read the original abstract

Noisy, partially overlapping data and the need for real-time processing pose major challenges for rigid registration. Considering that feature-based matching can handle large transformation differences but suffers from limited accuracy, while local geometry-based matching can achieve fine-grained local alignment but relies heavily on a good initial transformation, we propose a novel dual-space paradigm to fully leverage the strengths of both approaches. First, we introduce an efficient filtering mechanism consisting of a computationally lightweight one-point RANSAC algorithm and a subsequent refinement module to eliminate unreliable feature-based correspondences. Subsequently, we treat the filtered correspondences as anchor points, extract geometric proxies, and formulate an effective objective function with a tailored solver to estimate the transformation. Experiments verify our method's effectiveness, as demonstrated by a 32x CPU-time speedup over MAC on KITTI with comparable accuracy. Project page: https://ustc3dv.github.io/DualReg/.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper proposes DualReg, a dual-space paradigm for rigid registration of noisy and partially overlapping point clouds. It combines feature-based matching (which handles large transformations) with local geometry-based refinement (which achieves fine accuracy but needs a good initial guess). The pipeline first applies a lightweight one-point RANSAC filter plus refinement module to discard unreliable feature correspondences, then treats the surviving matches as anchor points, extracts geometric proxies, and solves a tailored objective function for the rigid transform. Experiments on KITTI are said to demonstrate a 32x CPU-time speedup over MAC while preserving comparable accuracy.

Significance. If the filtering step reliably produces usable anchors and the reported speedup/accuracy hold under realistic conditions, the method offers a practical efficiency gain for real-time robotics applications such as SLAM and object tracking. The hybrid design that leverages complementary strengths of feature and geometric matching is a sensible response to the stated challenges of noise and partial overlap.

major comments (2)
  1. [Experiments] Experiments section: the headline claim of a 32x CPU-time speedup over MAC on KITTI with comparable accuracy is presented without quantitative metrics, error bars, full baseline tables, or explicit data-exclusion rules. Because the central efficiency and accuracy assertions rest on these results, the current support cannot be evaluated.
  2. [§3.1] §3.1 (one-point RANSAC filtering): the method assumes this step can reliably remove unreliable feature correspondences so that the surviving matches serve as stable anchors for geometric-proxy extraction. No ablation studies, failure-case analysis, or quantitative characterization of inlier/outlier rates under KITTI noise and partial-overlap regimes are supplied; if the assumption fails, both the accuracy parity and the claimed computational advantage of the dual-space pipeline are at risk.
minor comments (2)
  1. [§3.2] Notation for the geometric proxy and objective function could be introduced with a short table or diagram to improve readability for readers unfamiliar with the dual-space formulation.
  2. [Abstract] The abstract states 'Experiments verify our method's effectiveness' but does not name the dataset or key quantitative outcomes; adding one sentence would make the contribution clearer at first reading.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments and the recommendation for major revision. We address each major comment below and describe the revisions planned for the next version of the manuscript.

read point-by-point responses

  1. Referee: [Experiments] Experiments section: the headline claim of a 32x CPU-time speedup over MAC on KITTI with comparable accuracy is presented without quantitative metrics, error bars, full baseline tables, or explicit data-exclusion rules. Because the central efficiency and accuracy assertions rest on these results, the current support cannot be evaluated.

    Authors: We agree that the current experimental presentation would benefit from additional quantitative detail to allow full evaluation of the claims. In the revised manuscript we will expand the Experiments section with complete baseline comparison tables (including all methods, runtime breakdowns, and accuracy metrics), error bars computed over multiple runs, and explicit statements of data exclusion criteria and evaluation protocol on KITTI. These additions will directly support the reported 32x CPU-time speedup and accuracy comparability. revision: yes

  2. Referee: [§3.1] §3.1 (one-point RANSAC filtering): the method assumes this step can reliably remove unreliable feature correspondences so that the surviving matches serve as stable anchors for geometric-proxy extraction. No ablation studies, failure-case analysis, or quantitative characterization of inlier/outlier rates under KITTI noise and partial-overlap regimes are supplied; if the assumption fails, both the accuracy parity and the claimed computational advantage of the dual-space pipeline are at risk.

    Authors: The referee is correct that the filtering step is a foundational assumption whose reliability should be characterized more explicitly. Although the end-to-end KITTI results already demonstrate that the surviving anchors enable accurate registration, we will add dedicated ablation studies, quantitative inlier/outlier rate tables before and after filtering, and a failure-case analysis under the noise and partial-overlap conditions present in KITTI. These new results will be placed in §3.1 and the Experiments section. revision: yes

Circularity Check

0 steps flagged

No circularity: algorithmic pipeline validated externally via benchmarks

full rationale

The paper describes a constructed dual-space registration pipeline (one-point RANSAC filtering followed by geometric-proxy refinement and solver). All performance claims rest on external experimental comparisons (e.g., 32x speedup on KITTI) rather than any quantity defined in terms of fitted parameters or self-referential predictions. No equations reduce to their inputs by construction, and no load-bearing self-citations or uniqueness theorems are invoked in the provided text. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The method rests on the standard rigid-transformation model and the assumption that feature correspondences can be filtered to yield usable anchors; no new physical entities are introduced and only typical algorithmic thresholds are expected.

free parameters (1)
  • RANSAC inlier threshold
    A threshold used inside the one-point RANSAC filtering stage to classify correspondences; its concrete value is not stated in the abstract.
axioms (1)
  • domain assumption The scene undergoes a rigid transformation between the two point clouds.
    Invoked implicitly by the registration objective and solver.

pith-pipeline@v0.9.0 · 5684 in / 1169 out tokens · 45393 ms · 2026-05-21T22:35:32.572467+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

52 extracted references · 52 canonical work pages

  1. [1]

    Method for registration of 3-d shapes,

    P. J. Besl and N. D. McKay, “Method for registration of 3-d shapes,” in Sensor fusion IV: control paradigms and data structures , vol. 1611. Spie, 1992, pp. 586–606

    work page 1992

  2. [2]

    Object modelling by registration of multiple range images,

    Y . Chen and G. Medioni, “Object modelling by registration of multiple range images,” Image and vision computing, vol. 10, no. 3, pp. 145–155, 1992

    work page 1992

  3. [3]

    Aa-icp: Iterative closest point with anderson acceleration,

    A. L. Pavlov, G. W. Ovchinnikov, D. Y . Derbyshev, D. Tsetserukou, and I. V . Oseledets, “Aa-icp: Iterative closest point with anderson acceleration,” in 2018 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2018, pp. 3407–3412

    work page 2018

  4. [4]

    A symmetric objective function for icp,

    S. Rusinkiewicz, “A symmetric objective function for icp,” ACM Trans- actions on Graphics (TOG) , vol. 38, no. 4, pp. 1–7, 2019

    work page 2019

  5. [5]

    Fast and robust iterative closest point,

    J. Zhang, Y . Yao, and B. Deng, “Fast and robust iterative closest point,”IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 44, no. 7, pp. 3450–3466, 2021

    work page 2021

  6. [6]

    Fast point feature histograms (fpfh) for 3d registration,

    R. B. Rusu, N. Blodow, and M. Beetz, “Fast point feature histograms (fpfh) for 3d registration,” in 2009 IEEE international conference on robotics and automation . IEEE, 2009, pp. 3212–3217

    work page 2009

  7. [7]

    Fast global registration,

    Q.-Y . Zhou, J. Park, and V . Koltun, “Fast global registration,” in Com- puter Vision–ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11-14, 2016, Proceedings, Part II 14 . Springer, 2016, pp. 766–782

    work page 2016

  8. [8]

    Learning compact geometric features,

    M. Khoury, Q.-Y . Zhou, and V . Koltun, “Learning compact geometric features,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2017, pp. 153–161

    work page 2017

  9. [9]

    Ppfnet: Global context aware local features for robust 3d point matching,

    H. Deng, T. Birdal, and S. Ilic, “Ppfnet: Global context aware local features for robust 3d point matching,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 195–205

    work page 2018

  10. [10]

    3dmatch: Learning local geometric descriptors from rgb-d reconstruc- tions,

    A. Zeng, S. Song, M. Nießner, M. Fisher, J. Xiao, and T. Funkhouser, “3dmatch: Learning local geometric descriptors from rgb-d reconstruc- tions,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2017, pp. 1802–1811

    work page 2017

  11. [11]

    Fully convolutional geometric fea- tures,

    C. Choy, J. Park, and V . Koltun, “Fully convolutional geometric fea- tures,” in Proceedings of the IEEE/CVF International Conference on Computer Vision, 2019, pp. 8958–8966

    work page 2019

  12. [12]

    Guaranteed outlier removal for point cloud registration with correspondences,

    ´A. Parra Bustos and T.-J. Chin, “Guaranteed outlier removal for point cloud registration with correspondences,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 40, no. 12, pp. 2868–2882, 2018

    work page 2018

  13. [13]

    The perfect match: 3d point cloud matching with smoothed densities,

    Z. Gojcic, C. Zhou, J. D. Wegner, and A. Wieser, “The perfect match: 3d point cloud matching with smoothed densities,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019, pp. 5545–5554

    work page 2019

  14. [14]

    Pointdsc: Robust point cloud registration using deep spatial consistency,

    X. Bai, Z. Luo, L. Zhou, H. Chen, L. Li, Z. Hu, H. Fu, and C.- L. Tai, “Pointdsc: Robust point cloud registration using deep spatial consistency,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2021, pp. 15 859–15 869

    work page 2021

  15. [15]

    A practical o(n2) outlier removal method for correspondence- based point cloud registration,

    J. Li, “A practical o(n2) outlier removal method for correspondence- based point cloud registration,” IEEE Transactions on Pattern Analysis and Machine Intelligence , vol. 44, no. 8, pp. 3926–3939, 2022

    work page 2022

  16. [16]

    Correspondence selection with loose–tight geometric voting for 3-d point cloud registra- tion,

    J. Yang, J. Chen, S. Quan, W. Wang, and Y . Zhang, “Correspondence selection with loose–tight geometric voting for 3-d point cloud registra- tion,” IEEE Transactions on Geoscience and Remote Sensing , vol. 60, pp. 1–14, 2022

    work page 2022

  17. [17]

    Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,

    M. A. Fischler and R. C. Bolles, “Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography,”Communications of the ACM, vol. 24, no. 6, pp. 381–395, 1981

    work page 1981

  18. [18]

    Napsac: High noise, high dimensional robust estimation-it’s in the bag,

    P. H. Torr, S. J. Nasuto, and J. M. Bishop, “Napsac: High noise, high dimensional robust estimation-it’s in the bag,” in British Machine Vision Conference (BMVC), vol. 2, 2002, p. 3. JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 13

    work page 2002

  19. [19]

    Graph-cut ransac,

    D. Barath and J. Matas, “Graph-cut ransac,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 6733–6741

    work page 2018

  20. [20]

    Compatibility-guided sampling consensus for 3-d point cloud registration,

    S. Quan and J. Yang, “Compatibility-guided sampling consensus for 3-d point cloud registration,” IEEE Transactions on Geoscience and Remote Sensing, vol. 58, no. 10, pp. 7380–7392, 2020

    work page 2020

  21. [21]

    Sac-cot: Sample consensus by sampling compatibility triangles in graphs for 3-d point cloud registration,

    J. Yang, Z. Huang, S. Quan, Z. Qi, and Y . Zhang, “Sac-cot: Sample consensus by sampling compatibility triangles in graphs for 3-d point cloud registration,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1–15, 2021

    work page 2021

  22. [22]

    Rlsac: Reinforcement learning enhanced sample consensus for end-to- end robust estimation,

    C. Nie, G. Wang, Z. Liu, L. Cavalli, M. Pollefeys, and H. Wang, “Rlsac: Reinforcement learning enhanced sample consensus for end-to- end robust estimation,” in Proceedings of the IEEE/CVF International Conference on Computer Vision , 2023, pp. 9891–9900

    work page 2023

  23. [23]

    Ransac back to sota: A two-stage consensus filtering for real-time 3d registration,

    P. Shi, S. Yan, Y . Xiao, X. Liu, Y . Zhang, and J. Li, “Ransac back to sota: A two-stage consensus filtering for real-time 3d registration,” IEEE Robotics and Automation Letters , 2024

    work page 2024

  24. [24]

    3d registration with maximal cliques,

    X. Zhang, J. Yang, S. Zhang, and Y . Zhang, “3d registration with maximal cliques,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2023, pp. 17 745– 17 754

    work page 2023

  25. [25]

    Fastmac: Stochastic spectral sampling of correspondence graph,

    Y . Zhang, H. Zhao, H. Li, and S. Chen, “Fastmac: Stochastic spectral sampling of correspondence graph,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 17 857–17 867

    work page 2024

  26. [26]

    MAC++: Going further with maximal cliques for 3d registration,

    X. Zhang, Y . Zhang, and J. Yang, “MAC++: Going further with maximal cliques for 3d registration,” in International Conference on 3D Vision 2025, 2025

    work page 2025

  27. [27]

    Sparse iterative closest point,

    S. Bouaziz, A. Tagliasacchi, and M. Pauly, “Sparse iterative closest point,” in Computer graphics forum , vol. 32, no. 5. Wiley Online Library, 2013, pp. 113–123

    work page 2013

  28. [28]

    Shot: Unique signatures of histograms for surface and texture description,

    S. Salti, F. Tombari, and L. Di Stefano, “Shot: Unique signatures of histograms for surface and texture description,” Computer Vision and Image Understanding, vol. 125, pp. 251–264, 2014

    work page 2014

  29. [29]

    Pointnet: Deep learning on point sets for 3d classification and segmentation,

    C. R. Qi, H. Su, K. Mo, and L. J. Guibas, “Pointnet: Deep learning on point sets for 3d classification and segmentation,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 652–660

    work page 2017

  30. [30]

    D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features,

    X. Bai, Z. Luo, L. Zhou, H. Fu, L. Quan, and C.-L. Tai, “D3Feat: Joint Learning of Dense Detection and Description of 3D Local Features,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 6359–6367

    work page 2020

  31. [31]

    Predator: Registration of 3D Point Clouds With Low Overlap,

    S. Huang, Z. Gojcic, M. Usvyatsov, A. Wieser, and K. Schindler, “Predator: Registration of 3D Point Clouds With Low Overlap,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 4267–4276

    work page 2021

  32. [32]

    Geotransformer: Fast and robust point cloud registration with geometric transformer,

    Z. Qin, H. Yu, C. Wang, Y . Guo, Y . Peng, S. Ilic, D. Hu, and K. Xu, “Geotransformer: Fast and robust point cloud registration with geometric transformer,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 8, pp. 9806–9821, 2023

    work page 2023

  33. [33]

    Bansac: A dynamic bayesian network for adaptive sample consensus,

    V . Piedade and P. Miraldo, “Bansac: A dynamic bayesian network for adaptive sample consensus,” in Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision , 2023, pp. 3738–3747

    work page 2023

  34. [34]

    Clipper: A graph-theoretic framework for robust data association,

    P. C. Lusk, K. Fathian, and J. P. How, “Clipper: A graph-theoretic framework for robust data association,” in 2021 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2021, pp. 13 828–13 834

    work page 2021

  35. [35]

    Clipper+: A fast maximal clique algorithm for robust global registration,

    K. Fathian and T. Summers, “Clipper+: A fast maximal clique algorithm for robust global registration,” IEEE Robotics and Automation Letters , 2024

    work page 2024

  36. [36]

    Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration,

    Z. Chen, K. Sun, F. Yang, and W. Tao, “Sc2-pcr: A second order spatial compatibility for efficient and robust point cloud registration,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 13 221–13 231

    work page 2022

  37. [37]

    Turboreg: Turboclique for robust and efficient point cloud registration,

    S. Yan, P. Shi, Z. Zhao, K. Wang, K. Cao, J. Wu, and J. Li, “Turboreg: Turboclique for robust and efficient point cloud registration,” 2025. [Online]. Available: https://arxiv.org/abs/2507.01439

    work page arXiv 2025

  38. [38]

    3dregnet: A deep neural network for 3d point registration,

    G. D. Pais, S. Ramalingam, V . M. Govindu, J. C. Nascimento, R. Chel- lappa, and P. Miraldo, “3dregnet: A deep neural network for 3d point registration,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , 2020, pp. 7191–7201

    work page 2020

  39. [39]

    Deep Global Registration,

    C. Choy, W. Dong, and V . Koltun, “Deep Global Registration,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020, pp. 2514–2523

    work page 2020

  40. [40]

    Robust outlier rejection for 3d registration with variational bayes,

    H. Jiang, Z. Dang, Z. Wei, J. Xie, J. Yang, and M. Salzmann, “Robust outlier rejection for 3d registration with variational bayes,” in Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2023, pp. 1148–1157

    work page 2023

  41. [41]

    HyperGCT: A Dynamic Hyper-GNN-Learned Geometric Constraint for 3D Registration,

    X. Zhang, J. Ma, J. Guo, W. Hu, Z. Qi, F. Hui, J. Yang, and Y . Zhang, “HyperGCT: A Dynamic Hyper-GNN-Learned Geometric Constraint for 3D Registration,” arXiv preprint arXiv:2503.02195 , 2025

    work page arXiv 2025

  42. [42]

    Hemora: Unsupervised heuristic consensus sampling for robust point cloud registration,

    S. Yan, Y . Wang, K. Zhao, P. Shi, Z. Zhao, Y . Zhang, and J. Li, “Hemora: Unsupervised heuristic consensus sampling for robust point cloud registration,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , June 2025

    work page 2025

  43. [43]

    TEASER: Fast and Certifiable Point Cloud Registration,

    H. Yang, J. Shi, and L. Carlone, “TEASER: Fast and Certifiable Point Cloud Registration,” IEEE Transactions on Robotics , vol. 37, no. 2, pp. 314–333, 2021

    work page 2021

  44. [44]

    A robust inlier identification algorithm for point cloud registration via ℓ0-minimization,

    Y . Jiang, X. Tang, C. Cheng, and Y . Yuan, “A robust inlier identification algorithm for point cloud registration via ℓ0-minimization,” in Advances in Neural Information Processing Systems , 2024

    work page 2024

  45. [45]

    Multisource forest point cloud registration with semantic- guided keypoints and robust ransac mechanisms,

    W. Dai, H. Kan, R. Tan, B. Yang, Q. Guan, N. Zhu, W. Xiao, and Z. Dong, “Multisource forest point cloud registration with semantic- guided keypoints and robust ransac mechanisms,” International Journal of Applied Earth Observation and Geoinformation , vol. 115, p. 103105, 2022

    work page 2022

  46. [46]

    Centralized ransac-based point cloud registration with fast convergence and high accuracy,

    K.-L. Chung and W.-T. Chang, “Centralized ransac-based point cloud registration with fast convergence and high accuracy,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing , vol. 17, pp. 5431–5442, 2024

    work page 2024

  47. [47]

    An improved ransac outlier rejection method for uav-derived point cloud,

    B. Salehi, S. Jarahizadeh, and A. Sarafraz, “An improved ransac outlier rejection method for uav-derived point cloud,” Remote Sensing, vol. 14, no. 19, p. 4917, 2022

    work page 2022

  48. [48]

    An improved ransac-icp method for registration of slam and uav-lidar point cloud at plot scale,

    S. Zhang, H. Wang, C. Wang, Y . Wang, S. Wang, and Z. Yang, “An improved ransac-icp method for registration of slam and uav-lidar point cloud at plot scale,” Forests, vol. 15, no. 6, p. 893, 2024

    work page 2024

  49. [49]

    Sorkine-Hornung and M

    O. Sorkine-Hornung and M. Rabinovich. (2017) Least-squares rigid motion using svd. [Online]. Available: https://igl.ethz.ch/projects/ ARAP/svd rot.pdf

    work page 2017

  50. [50]

    Spare: Symmetrized point- to-plane distance for robust non-rigid registration,

    Y . Yao, B. Deng, J. Hou, and J. Zhang, “Spare: Symmetrized point- to-plane distance for robust non-rigid registration,” arXiv preprint arXiv:2405.20188, 2024

    work page arXiv 2024

  51. [51]

    Preda- tor: Registration of 3d point clouds with low overlap,

    S. Huang, Z. Gojcic, M. Usvyatsov, A. Wieser, and K. Schindler, “Preda- tor: Registration of 3d point clouds with low overlap,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021, pp. 4267–4276

    work page 2021

  52. [52]

    Are we ready for autonomous driving? the kitti vision benchmark suite,

    A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the kitti vision benchmark suite,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2012, pp. 3354–3361

    work page 2012