How do I set up reality capture platform workflows for comparing as-built conditions against plans?

Here’s a practical way to set up RealityCapture / reality-capture-platform workflows for as-built vs. design (plans/BIM/CAD) comparison.

1) Define the comparison goal first

Decide what you need to measure:

• Surface deviation: scanned reality vs. model surfaces
• Dimension checks: lengths, heights, offsets, plumbness
• Progress verification: installed vs. planned scope
• QA/QC tolerances: pass/fail against allowable thresholds

This determines whether you should compare:

• point cloud to point cloud,
• point cloud to CAD/BIM model,
• or mesh to model.

2) Capture data with alignment in mind

For reliable comparison, your reality capture needs a strong reference system.

Best practice capture inputs

• Drone imagery for exteriors / large sites
• Terrestrial photos for interiors and detail
• LiDAR / laser scans for high-accuracy geometry
• GCPs / control points for georeferencing
• Target markers or surveyed checkpoints for registration verification

Tips

• Use consistent overlap and coverage
• Capture from multiple angles to reduce occlusion
• Include scale/control where possible
• Keep metadata: date, station locations, coordinate system, tolerances

3) Establish a common coordinate system

This is the most important setup step.

Your plans/model and your as-built capture must share:

• the same origin,
• same units,
• same rotation/orientation,
• same vertical datum if applicable.

Common approaches

• Survey control-based georeferencing
• Best-fit alignment only for rough checks
• Known project coordinates from civil/structural design files

If your plans are in CAD/BIM, ensure:

• model is cleaned up,
• correct units (mm vs in vs ft),
• model is positioned properly,
• any unused geometry is removed.

4) Process the reality capture data

In RealityCapture or similar platform:

1. Align imagery / scans
2. Generate dense point cloud
3. Create mesh if needed
4. Scale / georeference / register
5. Check residuals and control errors
6. Export deliverables

For comparison workflows, point clouds are often best

A dense point cloud usually gives the most direct and flexible comparison against design geometry.

5) Prepare the design model for comparison

Before importing plans/BIM:

• Simplify if necessary
• Remove temporary construction elements if not relevant
• Break large models into logical zones if the platform benefits from it
• Verify model is in the correct coordinate system
• Create comparison surfaces or sections if needed

Helpful formats

• OBJ / FBX / STL for meshes
• E57 / LAS / LAZ / XYZ / PLY for point clouds
• IFC / RVT / DWG depending on the platform

6) Align design and as-built data

There are usually two methods:

A. Control-based alignment

Use survey points / GCPs / known reference points to place both datasets.

Best when:

• accuracy matters,
• you need defensible QA/QC,
• the project is coordinated with survey control.

B. Best-fit registration

Match surfaces or point sets automatically.

Best when:

• you need quick checks,
• the reference model may have small offsets,
• absolute coordinates are less important than relative deviation.

Often you’ll use:

• best-fit for an initial alignment,
• then control points to validate or lock the transform.

7) Run deviation analysis

Once aligned, compare the as-built point cloud to the plan/model.

Typical outputs

• Heat maps of deviation
• Colorized distance-to-surface analysis
• Cross sections
• Cut/fill style reports
• Point-by-point residuals
• Tolerance pass/fail summaries

Common tolerances

Examples:

• ±5 mm for precise interior fit-out
• ±10–20 mm for structural work
• larger tolerances for earthworks / site grading

8) Validate with checkpoints

Don’t rely only on visual alignment.

Use independent check points to confirm:

• horizontal accuracy,
• vertical accuracy,
• overall RMS error,
• local distortion.

A good workflow includes:

• control points used to align,
• checkpoints reserved only for QA.

9) Package results for stakeholders

Deliver the results in a way that construction, design, and QA teams can use:

• Annotated screenshots
• PDF deviation reports
• Section cuts at critical locations
• Exported point clouds / models
• BIM issue references
• Dashboards if your platform supports them

10) Recommended workflow structure

A simple production workflow is:

1. Capture photos/scans
2. Register and georeference
3. Clean and segment point cloud
4. Import design model
5. Align to common coordinates
6. Run deviation analysis
7. Validate with checkpoints
8. Issue report and markups
9. Track corrections / re-capture if needed

Practical tips

• Use the same units everywhere
• Lock down coordinate systems early
• Segment by zones to reduce model size
• Keep a version history of both captures and plans
• Document what comparison method was used, since results vary by method
• If possible, define a tolerance matrix by element type

If you want the workflow in RealityCapture specifically

A typical RealityCapture setup would be:

• Import images or scan data
• Align
• Define control points / scale constraints
• Generate point cloud / mesh
• Export as point cloud
• Import the design model into a comparison tool
• Compute deviations and generate reports

RealityCapture is excellent for reconstruction, but many teams do the comparison step in tools like:

• CloudCompare
• Navisworks
• ReCap
• Trimble / Leica ecosystem tools
• BIM coordination software

If you want, I can also give you:

1. a step-by-step RealityCapture workflow,
2. a sample QA/QC checklist, or
3. a comparison workflow for BIM (Revit/IFC) vs scan data.