What Is LiDAR? Fundamentals for Builders and Architects

"LiDAR" is one of those words that come up more and more often in architects' offices, on building sites and in real-estate meetings — and that many people nod along to without quite knowing what is going on under the hood. This article explains LiDAR the way we experience it day to day: no marketing fluff, with the places where it really shines, and with the limits you should know before you place an order.

LiDAR — what the acronym actually stands for

LiDAR stands for Light Detection And Ranging — literally "detecting and measuring with light". The method is closely related to radar (which uses radio waves) and sonar (which uses sound); LiDAR uses laser light instead, typically in the near-infrared range. The scanner sends out very short laser pulses, measures the time until each reflected pulse returns, and converts that round-trip time into a distance. Do this thousands to millions of times per second across many angles and you get what is called a point cloud — a dense 3D set of measured points that together describe the geometry of a room or a landscape.

An important nuance: LiDAR is not a photographic process. A pure LiDAR point cloud has no colour information — only geometry and reflection intensity. That is why modern devices like the Matterport Pro3, which we use every day, combine the LiDAR measurement with HDR cameras. The end result is a colourised, photorealistic 3D capture you can actually walk through online.

How a LiDAR scan works, technically

At the heart of every LiDAR scanner sits a laser and a very precise timer. The physical principle is called Time-of-Flight: light travels roughly 30 cm in a nanosecond, so by measuring how long a pulse takes to return — and using the speed of light as a known constant — the device computes a very accurate distance.

Because a single point isn't enough, the scanner sweeps the laser beam through many different angles, either via a rotating mirror or via an arrangement of multiple lasers. On the Matterport Pro3 the scanner rotates once around its own axis during a capture and sweeps ceiling, walls and floor in a half-sphere. A typical scan position delivers several hundred thousand measurement points in 20 seconds.

Combining "round-trip time" with "current beam angle" yields one 3D point per pulse, expressed relative to the scanner's location. Once several scan positions have been captured with sufficient overlap, software merges the individual point clouds into a single global cloud in a step called registration. The overall quality of a capture is essentially decided here: well-chosen stations with proper overlap produce clean models, badly chosen ones lead to drift and gaps.

LiDAR vs photogrammetry vs structured light

LiDAR is not the only method that can capture a space in 3D. The three competing approaches you'll meet on a building project are:

In practice we often combine the methods: LiDAR for overall building geometry, drone photogrammetry for roofs and facades, and handheld structured-light scanners for spot details — for example a historical wrought-iron railing that needs to be documented at full resolution.

Range, accuracy and real-world limits

Range and accuracy come down to the device. For the Matterport Pro3, our default setup for indoor spaces and medium-sized outdoor areas, the figures are:

• Range: up to 100 m per laser pulse — large warehouses, industrial halls and outdoor facades can be captured without additional stations.
• Accuracy: ±2 cm per 10 m of distance. Across a 50 m hall the cumulative figure can settle around ±10 cm because registration errors from multiple stations add up.
• Point density: roughly 100,000 points per second — a single station produces more geometric data in 20 seconds than an architect could measure by hand in an entire day.

The honest limits of LiDAR are about materials, not devices. Three surface types give every LiDAR scanner trouble:

• Glass and mirrors: the laser either passes through (glass) or bounces off as if from a wall (mirror). In both cases we don't measure the actual surface. Workarounds: tag windows in advance, briefly cover mirrors, or correct the point cloud in post-processing.
• Deep black or strongly glossy surfaces: they absorb too much or reflect too little of the pulse. The Pro3 handles this better than older scanners, but a glossy black piano or matte-black industrial equipment may need extra scan stations.
• Moving objects: vehicles, people walking through, moving machinery create "ghosts" in the point cloud. If the building is in operation we plan scans so the area is briefly still.

Practical use cases on construction projects

LiDAR is not an end in itself. It pays off where conventional methods would be slow, risky or inaccurate. Five scenarios from our everyday work:

• Indoor spaces with little or no light: cellar vaults, plant rooms, service floors, freshly stripped shell buildings. Because LiDAR brings its own "light", we scan in complete darkness with the same precision as in daylight — photogrammetry would fail entirely here.
• Large halls and logistics areas: industrial halls, workshops and warehouses with 50–80 m spans are exactly where the Pro3's 100 m range pays off. Instead of fifty handheld measurement positions, 15–20 scan stations typically cover the whole building.
• Outdoor facades and courtyards: for an as-built model of an apartment building we combine ground-based LiDAR with drone photogrammetry for roof and upper floors. LiDAR ensures geometry is correct where the drone is not allowed to go.
• Tunnels, shafts and basements: classical GPS does not work there, and photogrammetry fails for lack of light. LiDAR is often the only viable method — tunnel sections, drinking-water reservoirs and underground car parks can be captured in a fraction of the time a traditional dimensional survey would take.
• Heritage preservation and archaeological documentation: for historic buildings the point cloud doubles as a digital archive and a planning basis. Should an element be damaged later or need restoration, an accurate 3D capture of the original state is already on file — practically unattainable at this density with traditional methods.

What you actually receive as a client

A useful LiDAR capture doesn't end with the raw scan. Our projects typically result in one of three delivery levels — which one is right depends on what you plan to do next:

• Walkable 3D tour: the most accessible form for builders, investors and marketing teams. Walk through it online in any browser, with a measuring-tape function. Sufficient for marketing, documentation and basic planning.
• Point cloud (E57, LAS, RCP): the raw data form — directly importable into Revit, ArchiCAD, AutoCAD and similar CAD/BIM tools. Architects, structural engineers and energy consultants normally work with this.
• BIM model or mesh: derived from the point cloud, with clean surfaces — walls, floors and ceilings as proper building components. The right choice when a planning application, a refurbishment design or a digital twin needs to come out of the data.

Which level is right for your specific project is something we clarify up front — a pure marketing tour does not need a BIM workup, while a refurbishment project in an existing building won't get far without CAD-ready data.

Bottom line and consultation

LiDAR is today the most pragmatic method for capturing buildings fast and accurately — from a single-family house through industrial halls to historic landmarks. It is not the highest-precision gold standard (terrestrial laser scanning from a survey office still owns that ground), but in terms of price-to-performance it delivers by far the best result for most construction and architecture use cases.

At Bitblade Vision we use the Matterport Pro3 as our default device and complement it project by project with drone photogrammetry or handheld scanners. Clearly bookable, transparent pricing, available across Germany and Switzerland.

If you're wondering whether LiDAR is the right choice for your project: a consultation costs nothing — and we'll tell you honestly if a different method would suit you better.