Semantic Segmentation of Flanges in Industrial Point Clouds Using PointNet++
DOI:
https://doi.org/10.5753/isys.2026.7126Keywords:
3D point cloud, Semantic segmentation, Scan-to-BIM, Industry 4.0, Information Systems, PointNetAbstract
Industry 4.0 has fostered the use of 3D point clouds for inspection, as-built modeling and decision support in engineering. However, the automated interpretation of such data remains challenging, particularly in complex industrial environments where the manual identification of components is time-consuming and error-prone. Although recent studies address the semantic segmentation of industrial scenes, none of them treats the flange as an isolated target class, which is the specific gap addressed here. This work investigates the use of PointNet++ for point-wise semantic segmentation of flanges in industrial equipment point clouds. We propose a pipeline that preserves physical scale (metric coordinates), computes geometric features with fixed physical radii, and samples spherical chunks of variable radius during training to capture multi-scale context. The network is trained with a combined Focal + Lovasz-Softmax loss and uses GroupNorm to stabilise inference with small batch sizes. On the test set, the final model achieved an overall accuracy of 90.41% and a mean IoU of 70.47%, with a recall of 90.75% and an F1 of 68.09% for the flange class. Comparative experiments against four traditional machine learning baselines (Random Forest, Gradient Boosting, Logistic Regression and XGBoost) demonstrated the clear superiority of the deep learning approach, which outperformed the best classical model by over 37 percentage points in mean IoU. An ablation study comparing three levels of data augmentation revealed that heavy domain randomisation can hurt performance in small, homogeneous datasets, motivating the adoption of a lightweight training pipeline. From an Information Systems perspective, the results demonstrate the feasibility of transforming raw scanning data into semantically enriched information capable of supporting Scan-to-BIM workflows, asset management and maintenance planning in industrial plants. The study also discusses organisational and socio-technical implications, as well as limitations related to dataset size and class imbalance.
Downloads
References
Abreu, N., Pinto, A., Matos, A., and Pires, M. (2023). Procedural point cloud modelling in Scan-to-BIM and Scan-vs-BIM applications: a review. ISPRS International Journal of Geo-Information, 12(7):260.
Autodesk (2024). Building information modeling (BIM). Disponível em: [link]. Acesso em: 23 maio 2025.
Badenko, V. et al. (2019). Scan-to-BIM methodology adapted for different applications. ISPRS - International Archives of the Photogrammetry Remote Sensing and Spatial Information Sciences, XLII-5/W2:1-7.
Becerik-Gerber, B. and Rice, S. (2010). The perceived value of building information modeling in the U.S. building industry. Journal of Information Technology in Construction (ITcon), 15:185-201.
Beraldin, J. A. et al. (2010). Advances in 3D imaging and modeling: acquisition, processing, and visualization. Remote Sensing, 2(8):2334-2388.
Berman, M., Triki, A. R., and Blaschko, M. B. (2018). The Lovász-Softmax loss: a tractable surrogate for the optimization of the intersection-over-union measure in neural networks. In 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 4413-4421.
Chen, J. et al. (2020). Integration of building information modeling and 3D point cloud data for augmented reality-based facility management. Automation in Construction, 119:103308.
Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13(3):319-340.
Géron, A. (2021). Hands-on machine learning with Scikit-Learn, Keras, and TensorFlow. O'Reilly Media, 3 edition.
Girardeau-Montaut, D. (2024). CloudCompare - open-source project. Disponível em: [link]. Acesso em: 20 set. 2025.
Glikson, E. and Woolley, A. W. (2020). Human trust in artificial intelligence: review of empirical research. Academy of Management Annals, 14(2):627-660.
Guo, Y. et al. (2016). A comprehensive performance evaluation of 3D local feature descriptors. International Journal of Computer Vision, 116(1):66-89.
Guo, Y. et al. (2021). Deep learning for 3D point clouds: a survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 43(12):4338-4364.
Jung, H. et al. (2021). Application of machine learning techniques in injection molding quality prediction: implications on sustainable manufacturing industry. Sustainability, 13(8):4120.
Lee, I. D., Lee, I., and Han, S. (2020). 3D reconstruction of as-built model of plant piping system from point clouds and port information. Journal of Computational Design and Engineering, 8(1):195-209.
Lin, T.-Y., Goyal, P., Girshick, R., He, K., and Dollár, P. (2018). Focal loss for dense object detection. arXiv preprint arXiv:1708.02002.
Liu, S. et al. (2021). 3D point cloud analysis: traditional, deep learning, and explainable machine learning methods. Springer Nature, Cham, Switzerland, 1 edition.
Maalek, R. et al. (2019). Extraction of pipes and flanges from point clouds for automated verification of pre-fabricated modules in oil and gas refinery projects. Automation in Construction, 103:150-167.
Matthews, C. (2023). The API ICP exam handbook: complete guide to passing the API 510/570/653 ICP exams. ASME.
Milletari, F., Navab, N., and Ahmadi, S.-A. (2016). V-Net: fully convolutional neural networks for volumetric medical image segmentation. arXiv preprint arXiv:1606.04797.
Noichl, F., Collins, F. C., Braun, A., and Borrmann, A. (2024). Enhancing point cloud semantic segmentation in the data-scarce domain of industrial plants through synthetic data. Computer-Aided Civil and Infrastructure Engineering, 39(10):1530-1549.
Poux, F. (2025). 3D data science with Python: building accurate digital environments with 3D point cloud workflows. O'Reilly Media, Inc.
Qi, C. R., Su, H., Mo, K., and Guibas, L. J. (2017). PointNet: deep learning on point sets for 3D classification and segmentation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pages 652-660.
Qi, C. R., Yi, L., Su, H., and Guibas, L. J. (2017). PointNet++: deep hierarchical feature learning on point sets in a metric space. CoRR, abs/1706.02413.
Rashdi, R., Martínez-Sánchez, J., Arias, P., and Qiu, Z. (2022). Scanning technologies to building information modelling: a review. Infrastructures, 7(4):49.
Remondino, F. and Campana, S. (2014). 3D recording and modelling in archaeology and cultural heritage - theory and best practices. BAR International Series 2598. Archaeopress.
Saranti, A., Pfeifer, B., Gollob, C., Stampfer, K., and Holzinger, A. (2024). From 3D point-cloud data to explainable geometric deep learning: state-of-the-art and future challenges. WIREs Data Mining and Knowledge Discovery, 14(6):e1554.
Sarker, S., Sarker, P., Stone, G., Gorman, R., Tavakkoli, A., Bebis, G., and Sattarvand, J. (2024). A comprehensive overview of deep learning techniques for 3D point cloud classification and semantic segmentation. Machine Vision and Applications, 35(4).
Shigeta, K., Nagumo, T., and Masuda, H. (2023). Point cloud segmentation for pipelines in industrial facilities using recurrent networks. Computer-Aided Design and Applications, pages 786-796.
Stojanovic, V., Trapp, M., Richter, R., and Döllner, J. (2019). Service-oriented semantic enrichment of indoor point clouds using octree-based multiview classification. Graphical Models, 105:101039.
Succar, B. (2009). Building information modelling framework: a research and delivery foundation for industry stakeholders. Automation in Construction, 18(3):357-375.
Valero, E. et al. (2021). An integrated scan-to-BIM approach for buildings energy performance evaluation and retrofitting. In Proceedings of the 38th International Symposium on Automation and Robotics in Construction (ISARC).
Venkatesh, V., Morris, M. G., Davis, G. B., and Davis, F. D. (2003). User acceptance of information technology: toward a unified view. MIS Quarterly, 27(3):425-478.
Wang, Q., Guo, J., and Kim, M.-K. (2019). An application-oriented scan-to-BIM framework. Remote Sensing, 11(3):365.
Wang, Q. and Kim, M.-K. (2019). Applications of 3D point cloud data in the construction industry: a fifteen-year review from 2004 to 2018. Advanced Engineering Informatics, 39:306-319.
Wu, Y. and He, K. (2018). Group normalization. In Proceedings of the European Conference on Computer Vision (ECCV).
Yin, C., Wang, B., Gan, V. J. L., Wang, M., and Cheng, J. C. P. (2021). Automated semantic segmentation of industrial point clouds using ResPointNet++. Automation in Construction, 130:103874.
Yin, C., Cheng, J. C. P., Wang, B., and Gan, V. J. L. (2022). Automated classification of piping components from 3D LiDAR point clouds using SE-PseudoGrid. Automation in Construction, 139:104300.
Zhao, Z., Zhang, Z., Nie, Q., Liu, C., Zhu, H., Chen, K., and Tang, D. (2024). Probing a point cloud based expeditious approach with deep learning for constructing digital twin models in shopfloor. Advanced Engineering Informatics, 62:102748.
Zhu, Q., Fan, L., and Weng, N. (2023). Advancements in point cloud data augmentation for deep learning: a survey. arXiv preprint arXiv:2308.12113.
Zoumpekas, T., Salamó, M., and Puig, A. (2022). Rethinking design and evaluation of 3D point cloud segmentation models. Remote Sensing, 14(23):6049.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 iSys - Journal of Information Systems
This work is licensed under a Creative Commons Attribution 4.0 International License.
