A point cloud is a 3D set of measured XYZ points that represents the surface geometry of a real-world object or site. Each point typically carries colour (RGB), an intensity value, and sometimes a classification (ground, building, vegetation). Point clouds are the raw material for BIM models, clash detection, deformation monitoring, façade documentation, heritage archiving, and any workflow that needs the existing geometry of the built or natural environment captured at scale.
There are three core capture methods. Each is the right tool for a different scenario, and the cost difference between them — and the difference in the resulting deliverable — is significant. This guide explains each method, the accuracy and density you can expect, and a decision framework for choosing the right one. If you need a survey delivered, see our point cloud survey service.
The three capture methods
Accuracies below are quoted at 1 sigma (68% confidence) and mapped against the RICS Measured Surveys of Land, Buildings and Utilities (3rd edition) accuracy bands every Angell Surveys deliverable references — the RICS document publishes corresponding 2 sigma (95% confidence) values at double the 1 sigma figure.
| Method | Plan accuracy (1σ) | RICS Band | Typical density (pts/m²) | Best for | Cost band |
|---|---|---|---|---|---|
| Terrestrial Laser Scan (TLS) | ±2–6 mm | A / B / C | 10,000+ | Engineering / heritage / BIM survey | Medium |
| UAV LiDAR | ±15–25 mm | D / E | 200–1,000 | Large-area topographic, vegetated corridor | Medium |
| UAV Photogrammetry | ±10–25 mm | D / E | 5,000–50,000 | Open-air sites with texture / orthomosaic | Lower |
Terrestrial Laser Scanning (TLS)
A static instrument mounted on a tripod that pans 360° horizontally and 270° vertically, firing several hundred thousand laser pulses per second. Each pulse returns a precise range and angle measurement, building a dense local point cloud in 5–15 minutes per setup. The instrument is moved to multiple scan stations across the site and the point clouds are registered — combined into a single coordinate frame — using shared targets, cloud-to-cloud algorithms, or surveyed station coordinates.
TLS is the gold standard for accuracy. Modern instruments (Leica RTC360, Faro Focus Premium, Trimble X9 and similar) achieve ±2 mm range noise at 25 m and densities exceeding 10,000 points per square metre. The full-colour imagery and HDR capture mean the resulting point cloud is also a visual record.
Used for: engineering survey of plant rooms / industrial facilities / mechanical and electrical first-fix, heritage façade documentation, BIM-handover scanning, mining gallery surveys, deformation monitoring of bridges and buildings. See our Highbury Stadium Square Estate facade case study for a worked example combining TLS, UAV photogrammetry and LOD 2–4 Revit modelling on the former Arsenal stadium converted to residential.
UAV LiDAR
A laser scanner mounted on a UAV, scanning the ground from above as the aircraft flies a planned pattern. UAV LiDAR units — Angell Surveys operates the Wingtra Ray with its LiDAR payload — typically achieve ±30–50 mm absolute accuracy with appropriate ground control, and crucially can capture bare earth under tree canopy and vegetation via multi-return processing — something photogrammetry simply cannot do.
UAV LiDAR is the right tool for large-area topographic surveys in vegetated terrain (woodland, scrubland, riparian zones), pipeline corridor mapping where ground levels matter beneath cover, and any application where ground exposure under canopy is critical. See our Calair Burn UAV LiDAR + photogrammetry case study for a 9.67 km upland catchment survey to a sub-50 cm DTM resolution supporting the Ballimore Restoration flood risk assessment.
It is more expensive per hectare than UAV photogrammetry, both because of the sensor cost and because the integrated INS/GNSS unit demands careful base-station setup. For open ground without vegetation, photogrammetry is usually the better value.
UAV Photogrammetry
Covered in depth in our drone photogrammetry guide. Briefly: hundreds of overlapping photos processed via Structure-from-Motion produce a dense point cloud — typically 5,000–50,000 points per square metre on textured surfaces. Plan accuracy of ±20–30 mm with proper Ground Control Points.
Photogrammetry’s strength is texture and density: the orthomosaic and RGB-coloured cloud are far richer than LiDAR returns. Its weakness is vegetation (cannot see through canopy) and low-light conditions.
A decision framework
Use this sequence:
- Is the target under tree canopy or dense vegetation? → UAV LiDAR
- Is the target GNSS-denied (tunnel, indoor)? → TLS
- Do you need sub-10 mm engineering-grade accuracy? → TLS
- Is the target open-air with hard detail and good lighting? → UAV Photogrammetry (best value)
- Combination of the above? → A multi-method capture is normal — TLS interior + UAV photogrammetry exterior + LiDAR on the vegetated boundary, all merged into one delivered point cloud.
File formats and delivery
Different downstream uses need different formats:
| Format | Strengths | Used by |
|---|---|---|
| LAS / LAZ | Open ASPRS standard, classifications supported, lossless compression (LAZ) | All major GIS / engineering software |
| E57 | Open ISO standard, preserves scan-station structure for TLS | TLS workflows, BIM authoring |
| RCP / RCS | Autodesk format (Recap point cloud) | Revit, AutoCAD, Civil 3D |
| PTS / PTX | Plain text — universal, large file size | Legacy and ad-hoc workflows |
| OBJ / FBX | Mesh derivatives, not raw point clouds | Visualisation, BIM coordination |
A common deliverable is LAS + E57 + RCP — LAS for any GIS or third-party engineering software, E57 for archive, RCP for direct Revit/Civil 3D ingest. The BIM team should specify the format up front; converting between them is straightforward but lossy in some directions (RCS doesn’t carry intensity or classification by default).
BIM workflow integration
For BIM scan-to-model work, the typical pipeline is:
- Capture — TLS for interior structure, UAV photogrammetry for exterior and roof, optionally UAV LiDAR for site context under vegetation.
- Register — combine all scan-station data into a single coordinate frame.
- Classify — separate ground, vegetation, walls, ceilings, MEP elements.
- Decimate — reduce density to BIM-workable size while preserving feature edges.
- Author — model in Revit using the classified point cloud as the reference, generating walls, slabs, columns, MEP runs at LOD 100–400 as required by the brief.
The point cloud is the as-built reality against which the BIM model is validated. A clash between the modelled element and the captured cloud is a real-world clash worth investigating.
What to put in a brief
- The capture method — or the accuracy/density/coverage requirement, leaving method selection to the surveyor
- The deliverable format(s) — be specific: LAS + E57 + RCP, with classifications X / Y / Z
- The reference system — OSGB36 / Newlyn ODN, or a contractor-defined site grid
- Coverage extent — drawn on a plan, not a verbal description
- Access window — when the site is available, and any safety constraints (live track, confined space, working at height)
- Required LOD if BIM authoring is included downstream
Frequently asked questions
Can I get a point cloud and a BIM model from the same survey? Yes — capture the point cloud, then author the BIM model from it. The order is fixed: capture first, model second. Modelling from imperfect capture costs more than recapturing.
How much storage does a typical point cloud need? A single TLS scan station produces ~1–2 GB. A medium site might be 20–50 scan stations = 30–100 GB. UAV LiDAR over a 50 ha site is typically 10–30 GB. Project early for storage and bandwidth — clients with on-premise BIM servers should confirm capacity before commissioning.
Can a point cloud be used for as-built documentation? Yes — that is one of its primary uses. A registered, classified point cloud captured at handover is a complete as-built record. Modern BIM contracts increasingly specify it as a deliverable alongside the model.
How long does the captured cloud remain useful? Indefinitely as a record of geometry at the time of capture. If the site changes (new construction, demolition, deformation), a fresh capture is needed. Change-detection between two captures is a common monitoring workflow.
Do you provide point cloud data only, or also extracted deliverables? Both. The raw cloud is one deliverable; extracted CAD/Revit elements, contour plans, sections and elevations are separate scopes that can be added.
For point cloud surveys across the UK using TLS, UAV LiDAR and UAV photogrammetry — single-method or hybrid — see our point cloud survey service. All deliverables ship in LAS, E57 and RCP as standard.