R3PM-Net: Real-time, Robust, Real-world Point Matching Network

2026-04-06

AI Summary

This work addresses the limited generalization and real-time performance of existing deep learning–based point cloud registration methods in real-world industrial settings, where noise, occlusion, and sparsity are prevalent. To overcome these challenges, we propose a lightweight, globally aware object-level point matching network that jointly optimizes accuracy and latency through robust feature extraction and an efficient correspondence filtering mechanism. The method is end-to-end trainable and achieves perfect registration on ModelNet40 within 7 ms (fitness = 1, inlier RMSE = 0.029 cm), offering approximately 7× speedup over RegTR. It also maintains high accuracy with inference latency under 50 ms on our newly introduced real-world datasets, Sioux-Cranfield and Sioux-Scans. Furthermore, this study presents the first efficient and robust registration framework tailored for practical industrial applications and publicly releases a novel dataset comprising photogrammetric and event-camera scans.

Abstract

Accurate Point Cloud Registration (PCR) is an important task in 3D data processing, involving the estimation of a rigid transformation between two point clouds. While deep-learning methods have addressed key limitations of traditional non-learning approaches, such as sensitivity to noise, outliers, occlusion, and initialization, they are developed and evaluated on clean, dense, synthetic datasets (limiting their generalizability to real-world industrial scenarios). This paper introduces R3PM-Net, a lightweight, global-aware, object-level point matching network designed to bridge this gap by prioritizing both generalizability and real-time efficiency. To support this transition, two datasets, Sioux-Cranfield and Sioux-Scans, are proposed. They provide an evaluation ground for registering imperfect photogrammetric and event-camera scans to digital CAD models, and have been made publicly available. Extensive experiments demonstrate that R3PM-Net achieves competitive accuracy with unmatched speed. On ModelNet40, it reaches a perfect fitness score of $1$ and inlier RMSE of $0.029$ cm in only $0.007$s, approximately 7 times faster than the state-of-the-art method RegTR. This performance carries over to the Sioux-Cranfield dataset, maintaining a fitness of $1$ and inlier RMSE of $0.030$ cm with similarly low latency. Furthermore, on the highly challenging Sioux-Scans dataset, R3PM-Net successfully resolves edge cases in under 50 ms. These results confirm that R3PM-Net offers a robust, high-speed solution for critical industrial applications, where precision and real-time performance are indispensable. The code and datasets are available at https://github.com/YasiiKB/R3PM-Net.

Problem

• Point Cloud Registration
• Real-time Performance
• Generalizability
• Industrial Applications
• Robustness

Innovation

• real-time point cloud registration
• robustness to real-world data
• lightweight deep learning
• global-aware matching
• industrial 3D vision