A point cloud is a set of measured points in 3D space, each with coordinates and often a colour or intensity value. Captured by laser scanners, mobile mapping or photogrammetry, a point cloud records the shape of an existing building or site in great detail. Scan-to-BIM is the process of modelling elements from that point cloud, so the existing conditions can be used in design, coordination and handover.
At a glance
| Captured with | Terrestrial laser scanners, mobile and handheld scanners, drones, photogrammetry |
| Size | Millions to billions of points; tens to hundreds of gigabytes for large sites |
| Open format | E57 (ASTM E2807); LAS/LAZ for aerial and infrastructure data |
| Common native formats | RCP/RCS (Autodesk), and each scanner vendor's own |
| Accuracy stated as | Level of accuracy (LOA), for example LOA30 |
From scan to model
- Plan the survey: scope, accuracy needed per area, scanner positions, control points.
- Capture: scans from many positions, each overlapping the next.
- Register: join the scans into one cloud, using targets or matching features, and tie it to the survey control so it sits in the project coordinates.
- Clean and segment: remove noise, people and vehicles; split by level or zone.
- Model: draw walls, slabs, structure and services against the cloud, to the agreed LOA and level of detail.
- Check: compare the model to the cloud and report deviations.
How point clouds are used
- Existing conditions for refurbishment: designers work against what is really there, not old drawings.
- Clash checking new work against existing: a new duct route is tested against the cloud or the scan model.
- Progress and quality checking: scans during construction are compared to the design model to find deviations early.
- As-built records: scans at completion support the record model.
- Heritage: detailed documentation of historic structures.
Using the cloud directly
Modern coordination tools can load point clouds next to models. For many checks, the cloud itself is enough and modelling everything is unnecessary. A good scope says which elements are modelled and which are left as points.
Common problems
- Registration not tied to survey control, so the cloud sits in its own coordinates.
- Scanner accuracy mistaken for model accuracy. The model is a simplification; state the represented LOA.
- Huge files shared without a plan: decimated or segmented clouds are easier to use.
- Hidden areas above ceilings and inside risers that were never scanned, modelled as if they were.
Where to read more
The USIBD Level of Accuracy Specification and national survey guidance (for example RICS guidance on measured surveys in the UK) cover specification and checking.