R-SLPR: Region-based Small-to-Large Point-cloud Registration with Contrastive Learning

2026-07-29

AI Summary

This work addresses the challenge of registering small-scale local point clouds to large-scale global reference point clouds, where significant scale disparity and minimal overlap render conventional registration methods ineffective. To overcome this, the authors propose a three-stage registration framework: first, discriminative local geometric patches are generated via Fibonacci grid partitioning and contrastive learning, with an explicit region proposal mechanism introduced to identify candidate matching regions; second, a region matching network performs coarse alignment; and finally, rigid transformation is refined through cascaded anchor selection and iterative optimization. Notably, this is the first approach to incorporate a region proposal mechanism into small-to-large point cloud registration, achieving a new state-of-the-art performance on ModelNet40, with mean absolute errors in translation and rotation reduced to 0.009 and 1.104, respectively—significantly outperforming existing methods.

Abstract

Point-cloud (PC) registration is fundamental to three-dimensional (3D) perception in robotic systems. However, classic registration algorithms falter when aligning a source PC containing limited, incomplete, or ambiguous geometric cues against a reference. This challenge of registering a small, partial PC to a significantly larger global reference is pervasive in real-world deployment yet remains insufficiently addressed by existing learning-based approaches, which typically assume comparable scales and significant overlap. To bridge this gap, we propose the Region-based Small-to-Large Point-cloud Registra- tion framework (R-SLPR), a novel three-stage architecture that fundamentally reformulates the scale-mismatched registration problem into a sequence of region proposal, regional matching, and iterative refinement. Unlike conventional methods that fail to localize specific regions, R-SLPR explicitly identifies candidate regions prior to estimating rigid transformations, ensuring robust alignment even under severe scale mismatch. The framework introduces a Fibonacci Grid Segmentation method coupled with a contrastive learning objective to effectively generate and match local geometric patches. Building on this, a novel Cascade Anchor Selection and Refinement algorithm iteratively aligns the source with the target region to maximize precision. Extensive evaluation on ModelNet40 demonstrates that R-SLPR establishes a new state-of-the-art accuracy standard, outperforming prior approaches and significantly reducing position and rotation Mean Absolute Error (MAE) to 0.009 and 1.104, respectively.

Problem

• point-cloud registration
• scale mismatch
• partial-to-global alignment
• 3D perception
• geometric ambiguity

Innovation

• point-cloud registration
• scale mismatch
• region proposal
• contrastive learning
• Fibonacci grid segmentation