Indoor Mapping Explained: LiDAR, Photogrammetry, and Mobile Scanning Compared

Indoor mapping is the process of turning a building's interior into a structured 3D dataset: walls, floors, columns, aisles and fixtures, captured with enough geometric accuracy that software can measure, navigate and update it later. It's the raw material behind every mall wayfinding app, every facility digital twin, and every BIM as-built drawing that actually matches the building instead of the architect's original intent.

For shopping malls and large commercial buildings, the practical choice usually comes down to three capture methods: LiDAR, photogrammetry, and mobile SLAM-based scanning that fuses elements of both. Each has a different accuracy profile, a different capture speed, and a different failure mode once you're standing in a dim parking garage or a mirrored ceiling. In the EU, and especially in France, Germany, Switzerland, Belgium and the Netherlands, the choice also intersects with data protection law more than it intersects with drone regulation, which surprises most people planning their first project.

The market underneath all this is growing fast. Global Market Insights values the LiDAR-in-mapping market at USD 5.3 billion in 2025, projecting growth to USD 45.8 billion by 2035 at a 25.6% compound annual growth rate, driven in part by integration with AI and digital twin platforms (GMI, accessed 2026). This piece compares the three methods on accuracy, speed and cost, walks through what actually happens on-site in a large retail building, and covers the regulatory and maintenance questions that decide whether a digital twin is still useful two years after it's made.

What an indoor mapping project actually produces

A finished indoor mapping job usually hands over three overlapping outputs: a point cloud (millions of individually measured 3D points), a floor plan or mesh derived from it, and a point-of-interest graph that wayfinding software actually reads. A digital twin, in this context, refers to a continuously updatable 3D digital replica of a physical space that mirrors its true current layout, not the architect's original drawings.

That distinction matters because malls change constantly. A kiosk moves. A wall goes up for a pop-up store. An escalator gets relocated during a renovation cycle. The CAD file handed over in 2019 stops being true the day the general contractor packs up, which is exactly the gap indoor mapping is meant to close.

LiDAR, photogrammetry, and mobile scanning, compared

LiDAR (Light Detection and Ranging) measures distance by timing how long a laser pulse takes to bounce off a surface and return to the sensor, building a dense point cloud without needing ambient light. Photogrammetry reconstructs 3D geometry from overlapping 2D photographs, using structure-from-motion algorithms to triangulate matched features across many images. Mobile SLAM scanning combines a LiDAR unit or depth camera with SLAM, short for simultaneous localization and mapping, which refers to the algorithm that lets a scanner build a map of an unknown space while simultaneously tracking its own position inside it. SurvTech Solutions notes that mobile scanning was largely limited to outdoor GPS-based positioning until SLAM matured enough to enable mobile indoor mapping without any satellite signal at all (SurvTech Solutions, accessed 2026).

Each method trades accuracy for speed and cost differently, and none of them is universally "best." The right one depends on what the twin is for.

Method	How it works	Typical indoor accuracy	Capture speed	Best fit	Main limitation
LiDAR	Laser time-of-flight, active sensor	Millimeter to low-centimeter (static/terrestrial)	Slower per station, very high point density	As-built BIM, MEP clash detection, benchmark surveys	Expensive hardware; no color without a paired camera; struggles on mirrors and glass
Photogrammetry	Structure-from-motion across overlapping photos	Centimeter-level, lighting-dependent	Very fast capture, slower processing	Texture-rich visual twins, leasing and marketing renders	Degrades on low-texture or dim surfaces; heavy compute load
Mobile SLAM scanning	LiDAR or depth camera plus SLAM localization, no GPS	Roughly 2-5 cm RMS over a full loop	Fastest for continuous walk-through capture	Whole-building capture: corridors, multi-level malls, warehouses	Drift accumulates over long loops without control points

How accurate is accurate enough

Accuracy requirements aren't fixed; they scale with what the output is used for. A wayfinding app needs positioning good to roughly a meter, not a centimeter. An as-built model feeding a BIM/facility management workflow needs centimeter-level fidelity so mechanical, electrical and plumbing runs actually line up. A leasing visualization needs visual richness more than metric precision, which is why photogrammetry and, increasingly, Gaussian Splatting show up in that use case even when a LiDAR pass exists for the geometry underneath.

A widely cited comparison published in Sensors evaluated three SLAM-based indoor mapping technologies against a reference 2D map using two feature-selection methods. The mapping RMS errors were 2.0 cm for SLAMMER, 3.9 cm for NavVis, and 4.4 cm for Matterport on interactively selected features, with corresponding detection rates of 100%, 98.9% and 92.3% (MDPI Sensors, accessed 2026).

System	RMS error (interactive features)	RMS error (MBR features)	Feature detection rate
SLAMMER	2.0 cm	1.7 cm	100%
NavVis	3.9 cm	3.2 cm	98.9%
Matterport	4.4 cm	4.7 cm	92.3%

The gap between these numbers is small in absolute terms but matters for procurement. A facility manager writing a spec that demands "survey-grade accuracy" for a wayfinding refresh is usually asking for hardware the project doesn't need, and paying for it.

Field reality: what scanning a mall actually looks like

Consider a mid-sized regional mall, roughly 45,000 square meters across three levels, with a glass atrium, an underground parking structure, and dozens of storefronts with mirrored or reflective finishes. A realistic capture plan splits the work: backpack or trolley-mounted SLAM scanning for the bulk of the corridors and common areas, done overnight or before opening hours to avoid foot traffic, plus targeted handheld photogrammetry passes for the atrium glass and any surface where lasers bounce unpredictably. The parking structure typically needs its own pass, since underground levels are GPS-denied by definition and have far less visual texture for photogrammetry to lock onto.

Teampl has run programs built around exactly this kind of operational complexity: mapping and scanning malls and commercial spaces across Germany, France, Benelux and Switzerland, collecting hundreds of scans a day for months, processed through a fully AI-optimized pipeline built specifically for that throughput across four markets. The bottleneck in projects like this is rarely the scanner. It's scheduling around retail hours, coordinating with multiple tenants who don't want their storefront geometry shared without sign-off, and getting a consistent data structure out of crews working in four countries with four sets of local practices.

A 2025 drone-based indoor mapping study, FlyMeThrough, illustrates the newer end of this spectrum: a commodity-drone system that produces 3D reconstructions with human-AI collaborative annotation of points of interest such as entrances, restrooms, stairs and elevators, evaluated across 12 indoor spaces with feedback from building managers and occupants (arXiv, accessed 2026). Drones make sense for large open volumes like atriums and warehouses; they make much less sense for narrow retail corridors where a person on foot with a backpack rig is faster and safer.

From raw scan to usable digital twin

Capture is the visible part of the job. The less visible part is the pipeline that turns raw point clouds into something a facility team or a navigation app can actually use, and it typically runs through four stages.

• Registration: aligning individual scans or photo sets into one coherent coordinate system, closing loops so a corridor that starts and ends at the same doorway actually matches up.
• Classification: separating the point cloud into walls, floors, ceilings, and movable fixtures, since a wayfinding graph doesn't need to know about a display rack that gets rearranged monthly.
• Floor plan and mesh extraction: generating the 2D or 3D representation that downstream software (BIM tools, wayfinding kiosks, leasing platforms) actually consumes.
• POI tagging: labeling entrances, restrooms, elevators, and tenant units, increasingly with human-in-the-loop AI assistance rather than fully manual tagging.

One visualization trend worth tracking rather than chasing: Gaussian Splatting has moved from research curiosity to a mainstream visualization layer in the geospatial industry, connecting the metric precision of point clouds with the visual expressiveness of imagery (Geo-matching, accessed 2026). It doesn't replace the underlying survey-grade geometry, but it's changing how digital twins get presented to non-technical stakeholders like leasing teams and marketing.

Regulation differs more than geometry does

The physics of LiDAR and photogrammetry don't change at a border. The legal obligations around what you do with the data absolutely do, and the EU is a useful place to start because five of its largest economies handle indoor capture slightly differently.

Market	Personal data in scans	Indoor drone flight	What's distinctive
France	GDPR via CNIL; faces and plates are typically blurred before storage or publication	Enclosed indoor flight generally falls outside DGAC airspace rules; venue consent is still required	CNIL has published specific guidance touching 3D capture and biometric-adjacent data
Germany	GDPR plus a strong tradition around an individual's right to their own image; state-level DPAs enforce actively	Enclosed flight sits outside federal aviation rules; local building and safety codes still apply	Default practice leans more toward anonymization than most EU peers
Switzerland	Revised Federal Act on Data Protection (FADP), not GDPR, but with broadly similar obligations	FOCA governs airspace; enclosed indoor flight is typically outside its scope	Non-EU, so GDPR doesn't apply directly, though multinational mall operators usually apply it anyway for consistency
Belgium	GDPR enforced via the Belgian Data Protection Authority (APD/GBA)	Enclosed flight sits outside Belgian civil aviation rules	Bilingual signage requirements complicate standardized POI naming across Flanders and Wallonia
Netherlands	GDPR enforced via Autoriteit Persoonsgegevens	Enclosed flight sits outside Dutch civil aviation rules	Stronger public-sector push toward open BIM/IFC data for large public-facing buildings

Outside the EU, the emphasis often flips. In the US, aviation rules (FAA Part 107) are the more visible constraint for drone-assisted exterior and rooftop capture, while indoor privacy obligations vary by state rather than existing as a single federal framework. Assume the EU's GDPR-first pattern doesn't automatically transfer to a US or APAC deployment without checking local rules first.

Keeping the twin current

A digital twin captured once and never revisited is a snapshot with a shelf life. Retail tenant mix turns over, seasonal pop-ups appear and vanish, and structural changes from renovations or fire-code updates make parts of the original scan wrong. The real cost of skipping maintenance isn't the mapping itself; it's what accumulates while nobody's tracking it, a pattern covered in more depth in Mapping Debt: The Bill Every Mall Eventually Pays. A practical rescan cadence for most malls sits somewhere between annual full rescans and targeted quarterly updates tied to known renovation schedules, rather than a single big-bang capture treated as permanent.

Positioning is a separate problem from the map

A perfect 3D model of a mall is useless for navigation unless something can tell a phone or kiosk where it currently is inside that model. That's a positioning problem, not a mapping problem, and it's usually solved with Wi-Fi RTT, Bluetooth Low Energy beacons, or visual positioning against the same imagery used to build the twin. This layer is what actually powers turn-by-turn directions once the underlying geometry exists, and it's covered separately in The Ambient Census: How Malls Actually Count You, which looks at how the same infrastructure doubles as foot-traffic analytics.

Where this breaks

None of these methods is close to solved in every condition, and it's worth stating the limits plainly rather than pretending otherwise. Mirrors and floor-to-ceiling glass confuse LiDAR by returning false or duplicate returns. Repetitive, texture-poor corridors, common in big-box interiors and parking structures, starve photogrammetry and SLAM of the visual features they need to localize accurately. Crowds during business hours make handheld and backpack capture slower and less repeatable, which is why most professional indoor scans happen before opening or after closing. And every method eventually needs some form of ground control or loop closure to stop cumulative drift from creeping into a map that's technically internally consistent but subtly wrong versus reality.

What to check before commissioning a scan

• Purpose first: wayfinding, BIM/facility management, and leasing visualization each demand a different accuracy tier; specifying survey-grade accuracy for a wayfinding refresh wastes budget.
• Access windows: confirm whether capture happens during closed hours, live during trading hours, or some mix, since this drives crew size and schedule.
• Data protection workflow: agree in advance how faces and plates get anonymized before storage, and who signs off on tenant-level storefront data.
• Update cadence: decide the rescan schedule before the first scan, tied to known renovation cycles rather than an arbitrary calendar date.
• Output format: confirm whether the deliverable is a raw point cloud, a classified mesh, an IFC/BIM model, or a simplified 2D floor plan, since each implies a different post-processing bill.

Start a project

Teampl runs field data collection programs, egocentric video, in-store audits, mystery visits, GIS surveys, designed around a specific research question, not a generic panel. If this topic touches a program you're planning, tell us what you're trying to learn.

Frequently Asked Questions