point cloud registration

This work addresses the challenging problem of rigid registration between partial and complete point clouds without initial pose estimates or feature correspondences. We propose the first direct semi-exhaustive search framework that performs global optimization jointly over rotation and translation spaces. Key methodological innovations include grid-based rotational sampling, analytical closed-form translation estimation, GPU-accelerated error evaluation, and an inlier-maximization strategy—enabling correspondence-free, end-to-end robust registration. On ModelNet40, our approach significantly outperforms existing state-of-the-art methods. In real-world robotic pose estimation tasks, it achieves high accuracy (mean rotation error < 1.5°, translation error < 0.02 m) and strong robustness against occlusion, noise, and low overlap ratios. This work establishes a new paradigm for unsupervised, globally optimal point cloud registration.

Addressing the challenge of simultaneously achieving robustness, efficiency, and generalization in global point cloud registration, this paper introduces an open-source C++ library. The proposed end-to-end pipeline integrates a lightweight Faster-PFH feature descriptor, a k-core graph-theoretic outlier pruning strategy, and robust pose solvers (e.g., RANSAC and TEASER+). Key contributions include: (i) Faster-PFH, which drastically reduces feature computation overhead while preserving discriminability; (ii) k-core pruning, lowering outlier rejection complexity from O(n²) to near-linear time; and (iii) a modular, highly extensible architecture that maintains high accuracy. Extensive experiments on standard benchmarks—including 3DMatch and KITTI—demonstrate that our method achieves 2–5× speedup over state-of-the-art robust registration approaches, with comparable registration accuracy, while supporting large-scale point clouds and cross-scenario generalization.

In point cloud registration, existing methods suffer from fixed iterative optimization paths, implicit correspondence refinement, and single-projection updates prone to local optima. This work introduces, for the first time, denoising diffusion models into the space of doubly stochastic matrices to explicitly model and optimize the distribution of matching matrices. Instead of fixed iterations, it employs the diffusion reverse process—enabling initialization from arbitrary inputs (e.g., white noise)—and integrates Sinkhorn regularization with differentiable geometric feature encoding to enable gradient-guided global matching search. Evaluated on 3DMatch/3DLoMatch and 4DMatch/4DLoMatch benchmarks, our approach achieves significant improvements in both rigid and non-rigid registration accuracy, correspondence quality, and robustness over RAFT-style methods and conventional feature-distance-based approaches.

In large-scale Structure-from-Motion (SfM), sparse inter-view overlap and drastic viewpoint changes—especially in aerial-to-ground scenarios—lead to low cross-image feature matching density and weak geometric consistency. To address this, we propose a geometry-guided hybrid matching paradigm: (1) geometric verification is formulated as an optimization problem based on Sampson distance; (2) detector-agnostic dense matching is fused with detector-driven sparse anchor guidance, where sparse anchors constrain and enhance the geometric consistency of dense matches; and (3) multi-view geometric consistency is explicitly modeled. Our method significantly improves both matching density and accuracy, outperforming state-of-the-art approaches in extreme large-scale settings. Consequently, camera pose estimation becomes more accurate, and the reconstructed 3D point cloud achieves higher completeness and fidelity.

Traditional point-to-point or point-to-plane distance metrics in non-rigid point cloud registration suffer from slow convergence and geometric detail loss. To address this, we propose a symmetric point-to-plane distance metric that jointly enforces positional and normal-based geometric constraints, significantly improving geometric fidelity. Methodologically, we introduce the first symmetric distance formulation for non-rigid registration and integrate it with a deformation-graph-based coarse alignment followed by an alternating optimization scheme within the Majorization-Minimization (MM) framework—balancing robustness, accuracy, and efficiency. Extensive experiments on multiple benchmark datasets demonstrate that our approach achieves state-of-the-art registration accuracy while maintaining high computational efficiency. The source code is publicly available.

This study addresses the trade-off in image–point cloud registration between insufficient inliers and an excessively high outlier ratio caused by suboptimal point cloud density, which limits registration accuracy. It presents the first systematic analysis of how point cloud density affects cross-modal registration and introduces a cross-coordinate correspondence pruning mechanism. Specifically, coarse correspondences are projected into the image coordinate system, where a lightweight network fuses geometric and feature information to predict inlier confidence scores for effective outlier rejection. Furthermore, a multi-density point cloud ensemble strategy is employed to enhance inlier recall. The proposed method consistently outperforms existing approaches across multiple benchmarks, achieving a registration recall improvement of at least 8.6%.

This work addresses the challenge of balancing accuracy and efficiency in 3D point cloud registration for remote sensing applications by proposing a fast registration algorithm based on probabilistic self-updating local correspondences and line vector sets. The method employs a dual-RANSAC interactive model to jointly optimize global and dynamically refined local correspondences, constructs a robust initial structure using angle histograms and line-length preservation, and incorporates a probabilistic self-updating mechanism along with a global early-stopping strategy to balance precision and computational cost. The optimal transformation is finally estimated via weighted singular value decomposition. Experimental results demonstrate that the proposed approach reduces root mean square error by at least 10% compared to state-of-the-art methods on public datasets while achieving significantly faster runtime.

This work proposes DC-Reg, a novel framework addressing the challenge of achieving globally optimal point cloud registration under partial overlap and large initial misalignment. The method introduces a unified difference-of-convex (DC) decomposition of the coupled objective function involving both transformation and correspondence variables, enabling the construction of a globally concave lower bound that significantly tightens the search space in branch-and-bound (BnB) optimization. By jointly exploiting the structural interdependence between transformation and correspondence variables, DC-Reg overcomes the limitations of conventional per-term relaxation strategies. Integrated with rotation-invariant features and an efficient solver for the linear assignment problem, the approach demonstrates faster convergence than existing global methods on both synthetic data and the 3DMatch benchmark, while exhibiting superior robustness under extreme noise and outlier conditions.

Establishing dense correspondences for 3D shapes in real-world scenarios is challenged by the absence of annotations, high resolution, topological distortions, and heterogeneous shape representations. This work proposes the ATM framework, which adopts a “model-then-match” paradigm by integrating pretrained vision foundation models with parametric shape priors to learn a unified shape representation in a shared parameter space from multi-view renderings. Dense correspondences are achieved in a zero-shot manner through geometric consistency constraints and spectral refinement, eliminating the need for correspondence-labeled training data. The method is inherently robust to topological noise and seamlessly handles diverse representations—including meshes, point clouds, and 3D Gaussians. It significantly outperforms existing approaches on non-isometric benchmarks, reducing correspondence errors by 73% on TOPKIDS and 37% on SMAL, while maintaining high efficiency and accuracy on wild scans with up to 200,000 vertices.