R3PM-Net: Real-time, Robust, Real-world Point Matching Network
Yasaman Kashefbahrami, Erkut Akdag, Panagiotis Meletis, Evgeniya Balmashnova, Dip Goswami, Egor Bondarau
2026-04-09
Summary
This paper introduces a new method, R3PM-Net, for accurately aligning 3D point cloud data, which is crucial for many applications like robotics and manufacturing.
What's the problem?
Existing methods for aligning 3D data often work really well in perfect, computer-generated environments, but struggle when dealing with real-world scans that are messy, incomplete, or have errors due to things like lighting or the scanning process itself. They also tend to be slow, which isn't ideal for applications needing quick responses.
What's the solution?
The researchers developed R3PM-Net, a faster and more robust system for aligning 3D point clouds. It's designed to handle imperfect data and still work quickly. To help test and improve their method, they also created two new datasets, Sioux-Cranfield and Sioux-Scans, containing realistic, flawed 3D scans. R3PM-Net works by intelligently matching features between the point clouds to figure out how they need to be rotated and translated to line up.
Why it matters?
This work is important because it provides a practical solution for aligning 3D data in real-world industrial settings where speed and accuracy are essential. R3PM-Net is significantly faster than previous methods while maintaining high accuracy, making it suitable for applications like quality control, automated inspection, and robotics where quick, precise 3D alignment is needed.
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 0.007s, 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.