Download PDFOpen PDF in browserDeep Learning Driven Framework for Construction Point-Cloud Processing10 pages•Published: June 2, 2026AbstractPoint-cloud data have become central to digital construction workflows through technologies such as laser scanning and photogrammetry. However, current processing methods remain fragmented, time-consuming, and heavily dependent on expert supervision. This study aims to clarify how construction workflows handle point-cloud processing, compare traditional and deep-learning-based approaches for key processing tasks, and propose a more cohesive workflow. A focused literature-based data collection process (2023–2025) identifies common processing tasks, denoising, sampling, registration, semantic segmentation, and completion, along with representative deep-learning applications and their advantages. The study summarizes these findings in an evidence table that links processing tasks, construction use cases, and task-level benefits of deep learning. Building on this analysis, it presents a conceptual framework that connects preprocessing and semantic inference into an integrated deep-learning-driven pipeline for construction point-cloud processing. The paper concludes by outlining research priorities for task-specific benchmarking, standardized datasets, and integration with construction information systems to enable reproducible and scalable automation.Keyphrases: construction automation, deep learning, point cloud, point cloud processing In: Wesley Collins, Anthony Perrenoud and John Posillico (editors). Proceedings of Associated Schools of Construction 62nd Annual International Conference, vol 7, pages 803-812.
|
