A point cloud is a set of measured points in 3D space. Each point has a position (an X, Y, and Z coordinate) and usually a colour value sampled from a camera. Put millions of them together and the surfaces of a room, a building, or a whole site emerge as a dense, measurable 3D record of how things actually are, not how a drawing says they should be.
Unlike a photo, a point cloud carries real dimensions. You can measure a doorway, check a floor-to-ceiling height, verify a beam clearance, or compare as-built conditions against design intent, directly from the data. That is why point clouds sit underneath most modern as-built documentation, renovation surveys, and Scan-to-BIM work: they replace tape measures and guesswork with a complete geometric record captured in a single visit.
How are point clouds captured?
There are three common ways to produce one, each with different trade-offs:
- Handheld SLAM LiDAR. You walk through the space and the scanner builds the cloud in real time using simultaneous localisation and mapping. Fast, flexible, and well suited to occupied or access-constrained sites: this is what the XGRIDS Lixel K2 does, capturing 200,000 points per second as you move.
- Terrestrial laser scanning. A tripod scanner such as the Leica BLK360 G2 captures dense, survey-grade points from fixed stations. Slower to set up, but the accuracy benchmark for contained, stable spaces. The trade-offs are covered in detail in handheld SLAM vs tripod scanner.
- Photogrammetry. Software reconstructs a cloud from many overlapping photos. No LiDAR required, but accuracy and scale depend heavily on capture discipline and lighting. See photogrammetry vs LiDAR vs 3DGS for the full comparison.
Whichever method captures the raw data, there is always a processing step between the device and a usable deliverable. For the XGRIDS scanners that step runs through LixelStudio on a workstation: raw capture in, registered and coloured point cloud out.
What file format should I ask for?
If you commission or produce a scan, the format question matters more than most people expect, because it decides which software can open the result. The formats you will meet:
- .las / .laz. The standard LiDAR exchange formats, maintained by the ASPRS. LAZ is a losslessly compressed LAS: same data, much smaller file. Ask for LAZ when transferring data and LAS when a tool insists on it.
- .e57. A vendor-neutral format that carries point data and imagery together. The safest answer when you do not yet know which software the downstream team uses.
- .rcp / .rcs. Autodesk ReCap projects, used to reference a cloud inside Revit, AutoCAD, and Navisworks. Usually produced by importing LAS/E57 into ReCap rather than exported directly from the scanner.
- .ply / .xyz. Simple, widely supported formats common in research and mesh tools, but they drop LiDAR-specific attributes such as intensity and classification.
A practical default: keep a LAZ master copy, hand E57 to anyone on unknown software, and let the BIM team build their own RCP from it.
How accurate is a point cloud?
Accuracy depends on the instrument, the capture technique, and how the result is registered, so a single number is never the whole story. Two numbers matter and they are often confused. Relative accuracy is how correct measurements are within the cloud: the K2 is specified at 1 cm, which is what renovation measurement and BIM modelling actually consume. Absolute accuracy is how well the cloud sits in a real-world coordinate system: with its built-in RTK the K2 reaches around 3 cm RMSE, which is what surveyors and site engineers care about when the data must line up with control or with other datasets.
Tripod scanners deliver millimetre-class relative accuracy and remain the benchmark for tolerance-critical work. Handheld SLAM trades a little precision for far greater speed; for most documentation, renovation, and mapping work, centimetre-level is more than the downstream deliverable needs. If the deliverable has a stated tolerance, agree it before capture and plan control points and verification measurements accordingly.
Point cloud vs mesh: which one do you need?
A point cloud is discrete samples; a mesh is a continuous surface of triangles built from those samples. The cloud is the more faithful record: nothing has been interpolated, so it is the right basis for measurement, deviation analysis, and BIM. A mesh is lighter to render and easier for visualisation, game engines, and 3D printing, at the cost of smoothing over fine detail and thin structures. The honest answer is that most projects want both at different stages, which is why the K2 workflow exports point cloud, mesh, and 3DGS from the same capture. Deliver the cloud to the engineers and the mesh (or a Gaussian-splat scene) to everyone who just needs to see the space.
From point cloud to BIM
Scan-to-BIM is the most common reason our renters capture point clouds. The workflow is consistent: capture the site, register and clean the cloud, bring it into Revit or Archicad as reference (usually via RCP or E57), then model walls, floors, structure, and services against the points. The cloud itself never becomes the model; it is the measurable evidence the model is checked against. AI-assisted tools now automate a useful share of the element extraction — see AI Scan-to-BIM for how that fits a rental workflow — but a human modeller still owns the result. Budget the modelling time separately from the capture: a site that scans in an hour can take days to model at LOD 300.
How much does a point cloud scan cost?
Commissioning a scanning firm is priced per project and varies with site size, deliverables, and tolerance requirements, so quotes differ widely. The alternative that Volumet exists for is renting the scanner and capturing the data yourself. The weekly dry-rental rates in Singapore (GST-exclusive): the handheld Lixel K2 at S$2,300, the Lixel K1 at S$2,200, the long-range Lixel L2 Pro at S$3,500, and the tripod Leica BLK360 G2 at S$2,800. Every first rental adds a one-time onboarding session: S$250 if you self-collect, S$400 with delivery to site.
The economics favour rental as soon as one week covers more than one site. A K2 week at S$2,300 spread across five captures is S$460 per site, with the data, the schedule, and the recapture decisions entirely in your hands. Availability and terms are on the rental page, or tell us your dates and site type and we will confirm a quote.