From Scan to Point Cloud — How We Process Data in Autodesk ReCap

Where We Left Off

In the previous article we looked at how the Leica BLK360 captures raw scan data on site — millions of measurement points from each scanner position, plus integrated HDR photography. But raw scan data on its own isn't a usable point cloud. Before it can be loaded into AutoCAD or Revit and used as a survey reference, the individual scans need to be processed and aligned in Autodesk ReCap.

This is the step that happens back in the office, and it's where the raw data becomes the clean, registered point cloud that your design team will actually work from.

What Is Autodesk ReCap?

Autodesk ReCap (Reality Capture) is the industry-standard software for importing, registering and managing laser scan data. It's part of the Autodesk ecosystem — which means it integrates directly with AutoCAD and Revit, making it the natural processing environment for survey data that will be used in those applications.

ReCap handles the entire post-processing workflow: importing raw scan files, registering scans to each other, cleaning the dataset, and exporting a final point cloud in a format that can be loaded directly into design software.

Step 1 — Importing the Raw Scans

The first step is importing the raw scan files from the BLK360 into a new ReCap project. Each scan position produces a separate file — a typical building survey might have anywhere from 15 to 100+ individual scan files depending on the size and complexity of the building.

ReCap reads these files and displays them as individual, unregistered point clouds — at this stage they're all in their own local coordinate systems and haven't been aligned to each other. They look like separate fragments of the building rather than a unified whole.

Step 2 — Scan Registration

Registration is the process of aligning all the individual scans into a single, unified coordinate system. This is the most technically important step in the workflow — a poorly registered point cloud will produce inaccurate drawings regardless of how carefully the scanning was done on site.

ReCap uses the overlapping geometry between adjacent scan positions to calculate how each scan relates to its neighbours. The software identifies common features — wall corners, floor edges, structural elements — that appear in multiple scans and uses these to compute the precise transformation needed to align them.

There are two main approaches to registration:

• Cloud-to-cloud registration — ReCap automatically matches overlapping areas between adjacent scans. This works well when there's sufficient overlap between positions and the geometry is complex enough to provide reliable reference points.
• Target-based registration — Reflective targets are placed in the building before scanning, and their known positions are used to anchor the registration. This is used when higher accuracy is needed or when scan overlap is limited.

Once registration is complete, ReCap reports a registration error — the average deviation between matched points across all scan pairs. For a good survey this should be in the range of 2–5mm. If any scan pairs show higher errors they're reviewed and corrected before proceeding.

Step 3 — Coordinate System Setup

Once the scans are registered to each other, the entire point cloud needs to be oriented correctly. For most building surveys this means:

• Setting the model to the correct orientation (aligning the building to true north or project north)
• Setting a meaningful datum level — typically finished floor level of the ground floor
• Where OS National Grid coordinates are required, geo-referencing the dataset using GPS control points captured on site

Getting the coordinate system right at this stage means that any drawings or models extracted from the point cloud will have the correct orientation and levels from the start — rather than needing to be rotated and adjusted later.

Step 4 — Cleaning and Filtering

Raw scan data always contains unwanted points that need to be removed before the dataset is finalised. Common sources of noise include:

• People and movement — anyone moving through the building during scanning appears as a smeared cluster of points
• Furniture and temporary objects — depending on the survey scope, loose furniture may need to be removed from the dataset
• Scanner self-reflections — highly reflective surfaces like mirrors and glass windows can produce spurious points
• Outdoor bleed — points captured through windows showing the outside environment

ReCap provides selection and deletion tools for manually removing these unwanted points, and automatic filters for removing isolated noise points that fall outside the expected geometry of the building.

Step 5 — Export

Once the registered and cleaned point cloud is ready, it's exported from ReCap in the appropriate format:

• RCP (ReCap Project) — the native Autodesk format, which loads directly into AutoCAD and Revit. This is the primary delivery format for most projects.
• E57 — an open standard format compatible with most other BIM and CAD applications including Bentley MicroStation, ArchiCAD and Trimble software.
• LAS / LAZ — a format used in GIS applications and for topographical data.

The exported RCP file is what gets loaded into AutoCAD or Revit for drawing production or BIM modelling — which is covered in the next two articles in this series.

How Long Does Processing Take?

Processing time depends on the number of scans and the complexity of the registration. A typical residential survey with 20–30 scan positions takes around 2–4 hours to process fully. A large commercial building with 80–100 scan positions might take a full day.

This processing time is built into our project programme — it's one of the reasons we quote 5–7 working days for standard turnaround rather than delivering the same day as the site visit.