Generalized-CVO: Fast and Correspondence-Free Local Point Cloud Registration with Second Order Riemannian Optimization

2026-06-08

AI Summary

This work addresses the challenges of low accuracy and efficiency in correspondence-free local point cloud registration, particularly in geometrically sparse scenes. The authors propose a continuous registration framework grounded in reproducing kernel Hilbert space (RKHS), where point clouds are modeled as continuous functions using anisotropic kernels that incorporate surface geometry. The method enforces tangential relaxation constraints and strengthens normal alignment, and—novelty—it introduces a second-order Riemannian optimization algorithm leveraging an approximate Riemannian Hessian for efficient solution. This approach significantly enhances convergence speed and robustness, reducing translational and rotational drift by over 55% in LiDAR and RGB-D tracking tasks. On object registration benchmarks, it outperforms ICP, achieving up to a tenfold speedup under moderate initial misalignments.

Abstract

We propose a fast and correspondence-free local point cloud registration method that leverages geometric surface structure and reproducing kernel Hilbert space (RKHS) embeddings. The method represents point clouds as continuous functions with point-wise anisotropic kernels that encode local geometry. This formulation improves alignment along surface normals while relaxing alignment along tangential directions. To solve the resulting registration problem, we propose a second-order on-manifold optimization scheme with approximate Riemannian Hessians, achieving a speedup of up to 10x over the first-order solvers used in prior correspondence-free RKHS-based methods. We demonstrate improved frame-to-frame LiDAR and RGB-D tracking accuracy across diverse indoor and outdoor datasets. On a LiDAR tracking registration task in the driving domain, we achieve a reduction of $>55\%$ in both translational and rotational drift in challenging feature-sparse environments. On object registration benchmarks, we show improved robustness over ICP-based methods and further gains when refining global initialization, particularly under moderate misalignment.

Problem

• point cloud registration
• correspondence-free
• local alignment
• LiDAR tracking
• RGB-D tracking

Innovation

• correspondence-free registration
• Riemannian optimization
• RKHS embedding
• anisotropic kernel
• point cloud alignment