FACT : Multinomial Misalignment Classification for Point Cloud Registration
(2025) 23rd Scandinavian Conference on Image Analysis, SCIA 2025 In Lecture Notes in Computer Science 15725 LNCS. p.324-337- Abstract
We present FACT, a method for predicting alignment quality (i.e., registration error) of registered lidar point cloud pairs. This is useful e.g. for quality assurance of large, automatically registered 3D models. FACT extracts local features from a registered pair and processes them with a point transformer-based network to predict a misalignment class. We generalize prior work that study binary alignment classification of registration errors, by recasting it as multinomial misalignment classification. To achieve this, we introduce a custom regression-by-classification loss function that combines the cross-entropy and Wasserstein losses, and demonstrate that it outperforms both direct regression and prior binary classification. FACT... (More)
We present FACT, a method for predicting alignment quality (i.e., registration error) of registered lidar point cloud pairs. This is useful e.g. for quality assurance of large, automatically registered 3D models. FACT extracts local features from a registered pair and processes them with a point transformer-based network to predict a misalignment class. We generalize prior work that study binary alignment classification of registration errors, by recasting it as multinomial misalignment classification. To achieve this, we introduce a custom regression-by-classification loss function that combines the cross-entropy and Wasserstein losses, and demonstrate that it outperforms both direct regression and prior binary classification. FACT successfully classifies point-cloud pairs registered with both the classical ICP and GeoTransformer, while other choices, such as standard point-cloud-quality metrics and registration residuals are shown to be poor choices for predicting misalignment. On a synthetically perturbed point-cloud task introduced by the CorAl method, we show that FACT achieves substantially better performance than CorAl. Finally, we demonstrate how FACT can assist experts in correcting misaligned point-cloud maps. Our code is available at https://github.com/LudvigDillen/FACT_for_PCMC.
(Less)
- author
- Dillén, Ludvig LU ; Forssén, Per Erik and Edstedt, Johan
- organization
- publishing date
- 2025
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- keywords
- Alignment Quality Prediction, Point Cloud Misalignment Classification, Point Cloud Registration, Regression-by-Classification
- host publication
- Image Analysis - 23rd Scandinavian Conference, SCIA 2025, Proceedings
- series title
- Lecture Notes in Computer Science
- editor
- Petersen, Jens and Dahl, Vedrana Andersen
- volume
- 15725 LNCS
- pages
- 14 pages
- publisher
- Springer Science and Business Media B.V.
- conference name
- 23rd Scandinavian Conference on Image Analysis, SCIA 2025
- conference location
- Reykjavik, Iceland
- conference dates
- 2025-06-23 - 2025-06-25
- external identifiers
-
- scopus:105009781600
- ISSN
- 0302-9743
- 1611-3349
- ISBN
- 9783031959103
- DOI
- 10.1007/978-3-031-95911-0_23
- language
- English
- LU publication?
- yes
- additional info
- Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.
- id
- 8e627e53-03fb-4b6e-b673-babd8e762b60
- date added to LUP
- 2026-01-09 17:19:32
- date last changed
- 2026-01-09 17:19:50
@inproceedings{8e627e53-03fb-4b6e-b673-babd8e762b60,
abstract = {{<p>We present FACT, a method for predicting alignment quality (i.e., registration error) of registered lidar point cloud pairs. This is useful e.g. for quality assurance of large, automatically registered 3D models. FACT extracts local features from a registered pair and processes them with a point transformer-based network to predict a misalignment class. We generalize prior work that study binary alignment classification of registration errors, by recasting it as multinomial misalignment classification. To achieve this, we introduce a custom regression-by-classification loss function that combines the cross-entropy and Wasserstein losses, and demonstrate that it outperforms both direct regression and prior binary classification. FACT successfully classifies point-cloud pairs registered with both the classical ICP and GeoTransformer, while other choices, such as standard point-cloud-quality metrics and registration residuals are shown to be poor choices for predicting misalignment. On a synthetically perturbed point-cloud task introduced by the CorAl method, we show that FACT achieves substantially better performance than CorAl. Finally, we demonstrate how FACT can assist experts in correcting misaligned point-cloud maps. Our code is available at https://github.com/LudvigDillen/FACT_for_PCMC.</p>}},
author = {{Dillén, Ludvig and Forssén, Per Erik and Edstedt, Johan}},
booktitle = {{Image Analysis - 23rd Scandinavian Conference, SCIA 2025, Proceedings}},
editor = {{Petersen, Jens and Dahl, Vedrana Andersen}},
isbn = {{9783031959103}},
issn = {{0302-9743}},
keywords = {{Alignment Quality Prediction; Point Cloud Misalignment Classification; Point Cloud Registration; Regression-by-Classification}},
language = {{eng}},
pages = {{324--337}},
publisher = {{Springer Science and Business Media B.V.}},
series = {{Lecture Notes in Computer Science}},
title = {{FACT : Multinomial Misalignment Classification for Point Cloud Registration}},
url = {{http://dx.doi.org/10.1007/978-3-031-95911-0_23}},
doi = {{10.1007/978-3-031-95911-0_23}},
volume = {{15725 LNCS}},
year = {{2025}},
}