Quick summary
- A terrestrial 3D laser scanner measures the distance to a surface and the horizontal and vertical angles of the beam, which places each measured point at an X, Y and Z coordinate. Millions of those points make up a point cloud.
- One scan only sees what is in line of sight, so a building is scanned from many stations. On our recent jobs that has ranged from about 85 positions for one production area to more than 350 in a stadium.
- The stations are joined (“registered”) into one point cloud, checked, and set to the agreed level, orientation and origin before anyone draws from it.
- Accuracy depends on range. High-end terrestrial scanners are specified at about 2 mm at 10 m, and on our projects points within about 10 m of a station on a well-registered cloud typically fall in the 2–4 mm range. Further away, uncertainty grows. The accuracy of the final drawings or model is specified separately.
- The point cloud is the measured dataset. Drawings, models and checks are produced from it when they are part of the agreed scope.
For how we scan, register and deliver a project, see our 3D laser scanning services.
What a laser scanner actually measures
This article is about terrestrial scanning of buildings and sites, the kind used for existing-conditions surveys. Handheld and object scanners often work differently. A terrestrial laser scanner sits on a tripod and sweeps a laser beam across everything around it. The head turns horizontally while a mirror sends the beam up and down, so a single setup covers almost the full sphere around the scanner, except the patch directly underneath it.
For each point the scanner records three things: the distance to the surface, and the horizontal and vertical angles the beam was pointing at. Distance plus the two angles places each point at a calculated coordinate in the scanner’s own coordinate system. Most scanners also record how strongly the light came back (intensity), and a built-in camera adds colour, so the point cloud looks like a photograph you can walk through and measure.
Time-of-flight and phase-shift
Scanners measure distance in one of two main ways. A time-of-flight scanner fires short pulses and times how long each takes to come back. A phase-shift scanner sends a continuous beam with modulated intensity and calculates distance from the phase difference between the transmitted and returned modulation (University of Arkansas CAST). Broadly, phase-based systems have been widely used for fast, dense building capture, while time-of-flight systems suit longer ranges. Modern scanners blur that distinction, so for existing-conditions work the more important questions are range, accuracy, density and how the scan network is set up: how many stations, how far apart, and how they are tied together.
How a building scan is captured, station by station
A scanner only sees what is in its line of sight. Columns, equipment and walls cast shadows where no points land, so the crew moves the scanner through the building and scans again, with enough overlap between stations for them to be joined later.
How many stations depends on how much is in the way:
- One production area in a food plant in Mississauga: about 85 positions in two days.
- A single-bay metal plant in Hamilton, about 280 by 50 feet: about 112 positions, with the columns and crane rails captured from both sides.
- Nine production lines in a food plant in Toronto: about 235 positions in two and a half days.
- The parapet and video board zone of the Rogers Centre: more than 350 positions.
Registration: turning many scans into one point cloud
Each station is captured in its own coordinate system, centred on the scanner. Registration finds how every station sits relative to the others and joins them into one point cloud. Software matches the overlapping geometry between stations, and targets placed on site (checkerboards or spheres) give fixed reference points where the geometry alone is not enough.
We register the stations and check their fit before setting the agreed project coordinates. For a local building grid, we orient the axes to agreed reference lines and set the requested origin; where survey control is required, we tie the dataset to that control. We then inspect the cloud in plan, section and clipping views to check that overlapping surfaces agree and there are no double surfaces. Levelling and orienting the cloud gives the team a usable coordinate framework. It does not flatten a floor or square a wall that is genuinely out of level or out of square: the measured slopes, offsets and out-of-square conditions stay in the data, because they are often the reason for the scan.
One lesson from our own workflow: we set alignment in the registration software, at the source, so every export carries the same coordinates. On one job, an attempt to realign a cloud in another program after registration came out badly off, and it had to be redone at the registration stage.
If the project needs survey coordinates, settle that before the scan. Targets that stay in place, identifiable and undisturbed, can let a surveyor determine their coordinates later and tie the cloud to the survey. If no usable reference points remain, establishing survey coordinates later may require a second visit. We explain why in 3D laser scanning survey: what you actually receive.
How accurate is it, and what affects accuracy?
Scanner accuracy is always stated at a range. A typical specification for a high-end terrestrial scanner is a 3D point accuracy of about 2 mm at 10 m and 3.5 mm at 25 m, measured at one sigma after warm-up and within the operating temperature range, with roughly 0.1 mm of extra uncertainty for every metre beyond 25 m (example manufacturer specification). That is where our working figure comes from: on our building projects, points captured within about 10 m of a station, on a well-registered cloud, typically fall in the 2–4 mm range. It is not a blanket guarantee for every point or dimension across a building. Accuracy depends on more than the specification sheet:
- Range. The further the surface, the larger the uncertainty and the wider the gap between points. Where accuracy matters, stations are planned so that key surfaces sit within about 10 m of the scanner.
- Angle of the beam. Surfaces hit at a shallow angle return weaker, noisier points than surfaces hit straight on.
- The surface itself. Dark surfaces return a weaker signal and noisier points than light ones, and shiny or transparent surfaces such as glass return unreliable points. Glass in particular is often captured by its frame rather than the pane.
- Registration. Registration error can accumulate through long or poorly constrained scan runs, which is why overlap, loop closure, targets or control matter more as a dataset grows.
Precision and accuracy are not the same thing
The two words are often used interchangeably, but in the lidar world they mean different things. The USGS lidar specification defines accuracy as the closeness of a measured value to an accepted true value, and precision (repeatability) as how closely measurements agree with each other, even if they all share the same bias (USGS Lidar Base Specification glossary). In practice, ASPRS describes lidar precision as the noise or “fuzziness” of the point cloud on a hard surface, measured without regard to survey control, while accuracy is measured against independent checkpoints (ASPRS, Photogrammetric Engineering & Remote Sensing).
- Precision shows up as how thin a flat wall or floor looks when you cut a section through the point cloud. Scanner makers state it as range noise, and it is far lower on light surfaces close to the scanner than on dark surfaces far away. Higher quality settings, which measure each point more times, reduce it.
- Relative accuracy is how well the points agree with each other across the whole dataset. That comes down to registration: whether a wall scanned from two stations lands in the same place.
- Absolute accuracy is how close the points are to their true position in a survey coordinate system. It can only be confirmed against independent control or checkpoints.
Which one matters depends on the deliverable. A slab elevation map or a clearance check depends mostly on precision and relative accuracy. A model that has to sit on a site survey also depends on absolute accuracy, which means control on site. A scan can be very precise and still be offset from the survey if it was never tied to control.
Resolution and quality: the two settings that matter on site
Two scanner settings decide what a station captures. Resolution sets how far apart the points land: the higher the resolution, the denser the points and the longer the scan. Spacing grows with distance, so a setting that gives a dense cloud at 5 m gives half that density at 10 m. Quality sets how many times each point is measured and averaged, which reduces noise at the cost of time. As a guide, manufacturers suggest lower resolutions for indoor spaces under about 10 m, medium resolutions for larger interiors, higher resolutions for outdoor work beyond 20 m, and a higher quality setting outdoors in sun or poor weather (example manufacturer guidance). In practice the right balance is the one that resolves the smallest thing the deliverable has to show, from the stations you can actually set up, in the time the site allows.
For drawings and models, scanner accuracy is only one part of the equation. The tolerance of the final deliverable should be agreed in the scope and checked with an appropriate QA method. The USIBD Level of Accuracy specification provides a framework for defining those requirements.
Terrestrial, mobile and 360° photos: what each is for
- Terrestrial scanning (on a tripod, station by station) gives the accuracy and density needed when something new has to fit something existing.
- Mobile scanning (handheld or carried) is much faster, but it can drift over distance unless tied to control. On a food plant job in Toronto, mobile was tried first and became a secondary dataset because it could not carry the precision the conveyors needed.
- 360° panoramas taken at each station are useful for seeing what something is and which way you are facing. The panorama image itself is not the measurement: dimensions should come from the registered point cloud, not from the photograph alone.
From point cloud to deliverables
The point cloud is the measured record of the building. Most teams do not design directly in it; they work from what is drawn or modelled from it, as agreed in the scope:
- The point cloud itself, as RCP or RCS for Autodesk software and E57 as the open archive copy. We cover which formats open where in point cloud file formats.
- 2D as-built drawings (floor plans, reflected ceiling plans, sections and elevations), traced from the cloud in AutoCAD.
- A Revit model to the agreed modelling scope and Level of Development (LOD), built against the points.
- Checks such as slab elevation and deviation maps, comparison with the design model, or clearance for new equipment. These are not the same as formal FF/FL floor flatness acceptance testing.
Frequently asked questions
How does a 3D laser scanner work?
It sweeps a laser beam around itself, measures the distance to each surface the beam hits, and records the horizontal and vertical angles of the beam. Distance plus those two angles gives each point an X, Y and Z coordinate, and millions of those points form a point cloud.
How long does it take to scan a building?
It depends on size and how crowded the space is. One production area took us about two days; nine production lines took two and a half. Drawings and models follow after registration.
How accurate is 3D laser scanning?
It depends on range. High-end terrestrial scanners are specified at about 2 mm at 10 m. On our building projects, points within about 10 m of a station on a well-registered cloud typically fall in the 2–4 mm range, with uncertainty growing at longer range. The tolerance of the final drawings or model is agreed separately.
What is the difference between precision and accuracy in laser scanning?
Precision is how tightly repeated measurements agree with each other, which shows up as noise on a flat surface. Accuracy is how close the points are to their true position, checked against independent control. A scan can be precise but offset, so a project that has to sit on survey coordinates needs control as well as a clean scan.
Is 3D laser scanning the same as LiDAR?
Terrestrial laser scanning is one form of LiDAR (light detection and ranging). The word LiDAR is also used for airborne, mobile and vehicle sensors, which are built for different jobs and accuracies.
Can a laser scanner see through walls or ceilings?
No. It only measures surfaces in its line of sight. Open ceiling tiles and access panels before the scan if concealed services matter.
What do I get at the end?
A registered point cloud, plus the drawings, models or checks included in the agreed scope: 2D as-built drawings, a Revit model, or deviation and slab elevation reports.
The scan is the measurement
Laser scanning replaces tape and memory with a measured record of the building as it stands, including the parts that are out of level, out of square or not where the drawings say. The scanner does the measuring; the quality of the result comes from the setup: enough stations, good control, careful registration and a deliverable with an agreed tolerance, matched to the question your team is asking.
