Fast-Asset: A Real Time Scan-To-3D Point Cloud BIM-Ready Asset Generation Method for Small MEP Elements
DOI: 10.35490/EC3.2026.204
Abstract: Building Information Modeling (BIM) requires timely as-is updates, but existing scan-to-BIM pipelines based on laser scanning or photogrammetry demand specialised hardware and offline processing, limiting frequent small-asset updates in digital-twin maintenance. This paper introduces Fast-Asset, a real-time on-site framework integrating mixed-reality capture, learning-based detection, and incremental 3D reconstruction to generate BIM-ready, pre-segmented point clouds of small MEP assets in a single walkthrough. Instant visual feedback enables in-situ validation and re-capture. Experiments demonstrate a 3-to-10x reduction in capture-to-model time versus laser scanning and photogrammetry, with comparable surface roughness and low background contamination, collapsing multi-stage offline processes into a single field operation.
Keywords: 3D Point Cloud Segmentation, Building Information Modeling (BIM), Object detection, Real-time 3D Capture, Scan-to-BIM