DualReg: Dual-Space Filtering and Reinforcement for Rigid Registration

TL;DR

DualReg combines feature-based and geometry-based methods in a dual-space paradigm with filtering and reinforcement to improve rigid registration accuracy and speed, especially for noisy and partial data.

Contribution

It introduces a novel dual-space approach with efficient filtering and a tailored solver, significantly enhancing registration speed and robustness over existing methods.

Findings

Achieves 32x CPU-time speedup over MAC on KITTI dataset.

Maintains comparable accuracy with improved efficiency.

Effectively filters unreliable correspondences for better registration.

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/.