Point Cloud Processing: From Raw Scans to Usable Data

Registration, cleaning, segmentation and format conversion — what happens between the field scan and your CAD or BIM workflow.

The point cloud you receive from a 3D scanning service has already gone through several processing steps. Understanding them helps you specify what you actually need and avoid paying for work you don't.

Step 1 — Registration

Each scan station captures a partial view. Registration aligns all stations into a single coordinate system. Two approaches:

• Cloud-to-cloud — software (Cyclone Register 360, FARO Scene, RealityCapture) matches overlapping geometry between scans. Fast for indoor work with clear features.
• Target-based — physical targets are placed in the scene and used as reference points. More accurate, required for surveying-grade projects or featureless environments (long corridors, blank facades).

Quality check: a good registration delivers mean error under 5 mm across the whole project.

Step 2 — Georeferencing (optional)

If your project needs to align with cadastral data or other geospatial sources, the registered cloud is georeferenced to a local coordinate system (e.g. WGS84 / UTM 36N for Cyprus). This adds 1–2 days of fieldwork plus GNSS or total station equipment.

Step 3 — Cleaning

Raw clouds contain noise: people walking through the scan, moving cars, vegetation outside windows, reflections. Cleaning removes these so downstream modeling isn't confused.

This step is often underestimated. For a busy commercial building, cleaning can take as long as the field scanning itself.

Step 4 — Segmentation and classification

Optional but valuable for large projects. Points are grouped by what they represent: ground, walls, vegetation, structures, MEP. Makes later modeling much faster and enables automated workflows.

Step 5 — Format conversion

Final deliverables depend on what software your team uses:

What to ask for

When commissioning a scan, specify upfront:

1. Target registration accuracy (typically 3–5 mm)
2. Whether georeferencing is needed
3. Required deliverable formats (we recommend e57 + rcp as a baseline)
4. Whether you need the raw cloud or also intensity / RGB data
5. Storage and handover method (point clouds can easily reach 50–200 GB)

A note on storage

A full scan of a multi-story office can produce 100+ GB of data. Plan for transfer via external SSD or large cloud storage — sending it over email won't work. We hand over data on encrypted SSDs or via secure WeTransfer / cloud links.

Need clean, processed point cloud data for your project in Cyprus? Get in touch.