Point Cloud Model: Technical Reference
This reference summarizes what a point cloud model contains, how it differs from a 3D model, and how the conversion pipeline is organized.
Definition
A point cloud model is a dataset of 3D coordinates that records the surface of a real object or space. Every point holds an X, Y and Z position and often colour and intensity values. The data comes from laser scanning (LiDAR) or photogrammetry. It is measured, but it has no structure: software reads coordinates, not walls or pipes.
Point attributes
| Attribute | Records | Use in conversion |
|---|---|---|
| XYZ coordinates | Exact position in space | Geometric skeleton for tracing |
| RGB colour | Photographic colour value | Tells one surface from another |
| Intensity | Strength of the laser return | Separates materials, such as metal duct and plaster |
| Surface normal | Direction the surface faces | Helps detect planes and clean edges |
| Timestamp | When the point was recorded | Aligns points from separate sessions |
Point cloud vs 3D model
| Point cloud | 3D model | |
|---|---|---|
| Form | Disconnected measured points | Organized geometry |
| Content | Coordinates only | Defined objects such as walls and floors |
| Actions | View and measure | Select, schedule, clash-check, edit |
A conversion produces either a mesh (a continuous surface, good for visualization, heritage shapes and VR) or a parametric model (objects that carry data, needed for engineering, coordination and quantities).
Pipeline
- Capture: a scanner or photogrammetry rig records distances to surfaces.
- Registration: separate scans are aligned into one cloud in a shared coordinate system and cleaned of noise.
- Conversion: the cloud is imported into authoring software, traced and classified, then checked against the source cloud.
Input quality conditions
- Point density must resolve the smallest required element.
- Coverage must reach every surface from enough angles.
- Registration accuracy must keep walls from doubling or ghosting.
- Noise must be filtered out so the true surface stays readable.
Level of Development
| LOD | Carries | Typical use |
|---|---|---|
| 200 | Approximate geometry | Early coordination and massing |
| 300 | Accurate geometry and dimensions | As-built documentation and design |
| 400 | Fabrication-level detail and data | Fabrication and clash-free MEP |
Tooling
Registration: Autodesk ReCap (RCP or RCS files) and Leica Cyclone. Authoring: Revit, ArchiCAD, Tekla Structures, plus plugins such as Undet that bring the cloud into Revit or SketchUp.
Applications
As-built documentation, renovation and retrofit design, construction verification (scan-vs-BIM), facility management, heritage preservation, earthworks and civil work, and structural, MEP and industrial plant modeling.
Terminology
LiDAR is a scanning technology. A point cloud is the dataset it produces. "Modeling" and "modelling" name the same process in American and British spelling.
Reference: https://vibimglobal.com/blog/point-cloud-model/
See more: