Handheld and SLAM Mobile Scanning for Heritage Documentation

What SLAM Scanning Is

SLAM stands for Simultaneous Localisation and Mapping — a technique, borrowed from robotics, by which a moving device builds a map of an unknown space while at the same time working out its own position within that map. A handheld SLAM scanner applies this to survey: as the operator carries it through a building, it continuously measures the surroundings and continuously computes where it is, stitching the measurements into a single coherent point cloud without any fixed setup positions.

The devices most heritage practitioners encounter are handheld or backpack LiDAR units (the Leica BLK2GO, GeoSLAM ZEB series and similar), which combine a spinning laser rangefinder with an inertial measurement unit and sometimes cameras. Because the scanner tracks its own motion, there is no separate registration step for the operator to perform — the walk-through produces an already-assembled cloud, which is a large part of the time saving.

How Handheld SLAM Works

A handheld SLAM scanner fuses two streams of data in real time. A laser scanner (usually a rotating LiDAR) measures distances to surfaces all around the device, thousands of points per second. An inertial measurement unit (IMU) tracks the device's acceleration and rotation. The SLAM algorithm combines these — using the geometry it is seeing to correct the drift of the inertial tracking, and the inertial tracking to know how the geometry from one instant relates to the next — to place every measured point into one consistent coordinate frame.

The critical consequence is that the scanner's accuracy depends on how well it can track its own trajectory, and that tracking relies on continuously seeing enough stable, distinctive geometry to correct itself. In feature-rich, well-connected spaces the algorithm stays well-constrained; in long, featureless or repetitive spaces — a plain corridor, a bare hall — the trajectory estimate drifts, and that drift shows up as accumulated positional error along the path.

The Accuracy Reality

The honest headline is that handheld SLAM is roughly an order of magnitude less accurate than terrestrial laser scanning. Typical SLAM relative accuracy is in the range of 20–60mm, and it can be worse over long, poorly-constrained paths; a good tripod TLS achieves a few millimetres. This is not a defect to be processed away — it is inherent to estimating position from motion rather than measuring from a fixed, levelled instrument.

That accuracy is entirely adequate for many heritage purposes and entirely inadequate for others. For an overall building record, a circulation study, a volume, a floor plan, or a rapid record of a space before intervention, 20–60mm is fine. For a 1:50 measured elevation of a carved facade, for structural deformation monitoring, or for anything requiring millimetre fidelity of fine ornament, it is not. Matching the method's accuracy to the required output accuracy is the entire decision.

SLAM vs TLS vs Photogrammetry

The three methods are not competitors so much as tools for different jobs. Terrestrial laser scanning gives the highest geometric accuracy and detail but is slow. Photogrammetry gives excellent detail and colour on textured surfaces at moderate speed and low equipment cost, but struggles in the dark and on featureless surfaces. SLAM gives fast, effortless coverage of complex and connected spaces at lower accuracy. The right choice — often a combination — follows from what the project actually needs.

Where SLAM Wins for Heritage

SLAM earns its place wherever speed and coverage of complex space matter more than millimetre accuracy. In heritage documentation, that describes a surprising number of real situations.

• Rapid whole-building capture — recording an entire structure's geometry and layout quickly, as context or as a base record.
• Complex, connected interiors — temple circulation, multi-room palaces, forts and monasteries where TLS would need dozens of setups to reach every space.
• Cluttered or occupied spaces — active temples, museums and occupied historic buildings where a walk-through captures around obstructions a fixed scanner would be blocked by.
• Emergency and pre-intervention records — capturing maximum geometry in minimum time before repair, demolition or disaster response, when nothing may be better than a fast complete record.
• Dark interiors — as an active LiDAR method, SLAM works in darkness where photogrammetry cannot.
• Floor plans and volumes — generating quick, adequately-accurate plans and space measurements for management and planning.

Where SLAM Falls Short

The limits follow directly from the method. SLAM should not be relied on where its accuracy or detail is inadequate, and pretending otherwise produces records that look complete but cannot support the work they are meant to.

• Measured drawings at 1:50 or finer — the 20–60mm accuracy does not meet the tolerance; use TLS or photogrammetry.
• Fine carved detail — SLAM point density and accuracy are too coarse to record ornament, inscriptions or tool marks faithfully.
• Deformation monitoring — detecting millimetre movement between surveys requires the repeatable precision SLAM does not have.
• Long, featureless spaces — bare corridors and plain halls starve the SLAM algorithm of the features it needs to track, causing drift.
• Surveys requiring a certified accuracy statement to a tight tolerance — SLAM's accuracy is real but modest, and must be honestly stated, not implied to be scan-grade.
• Colour-critical records — many SLAM units capture geometry with limited or no photographic colour compared with photogrammetry's photorealistic texture.

Capture Technique for Good SLAM Data

SLAM data quality depends heavily on how the operator moves. The algorithm needs a smooth, steady trajectory through feature-rich space, with loops that let it correct drift. Rushed, jerky, or straight-line-and-never-return captures produce the worst results.

1. 1Plan a route that forms closed loops returning to the start, rather than out-and-back lines, so the system can close loops and correct drift.
2. 2Walk smoothly and steadily at a moderate pace — sudden movements and rapid rotations degrade the trajectory estimate.
3. 3Keep feature-rich geometry in view; when passing through a long featureless corridor, slow down and let the scanner see the ends and any distinctive features.
4. 4Overlap your path — revisit junctions and pass through key spaces more than once to reinforce loop closure.
5. 5Avoid moving people and objects in the immediate capture where possible; dynamic clutter adds noise and can confuse tracking.
6. 6Where survey-grade tie is needed, place and capture a few control targets that can be measured independently, so the SLAM cloud can be checked or constrained against known points.

SLAM in the Indian Heritage Context

For Indian heritage sites, SLAM's speed suits several common constraints. Active temples with continuous worship and footfall allow only limited, non-disruptive access — a quiet early-morning walk-through captures far more than a day of tripod setups would in the same window. Sprawling complexes with many connected mandapams, corridors and enclosures are exactly the connected-space case SLAM handles best. And where documentation must be done quickly ahead of a festival, a repair or a monsoon, a rapid complete record has real value.

The same permission framework that governs any survey on a protected monument applies to SLAM: access to document an ASI-protected or state-protected monument requires the appropriate permission, and any drone-mounted variant additionally requires drone clearance. SLAM's advantage under these constraints is time — obtaining permission for a short, non-invasive walk-through is often simpler than for extended tripod-based work, and the capture itself disturbs the site far less.

Common Mistakes

• Using SLAM where the output needs 1:50 or finer accuracy — the method's 20–60mm accuracy cannot support it; use TLS or photogrammetry.
• Walking straight-line, never-return paths — without loop closure, drift accumulates uncorrected along the path.
• Moving too fast or jerkily — rapid motion degrades the trajectory estimate and the whole cloud's quality.
• Capturing long featureless corridors at speed — starves the algorithm of tracking features and causes drift; slow down and capture the ends.
• Presenting a SLAM cloud as a precise survey — its visual completeness invites overconfidence; state the real accuracy with the deliverable.
• Expecting fine carved detail — SLAM is too coarse for ornament and inscriptions; add photogrammetry or TLS for detail.
• Skipping independent check points where accuracy matters — without a few known points, the SLAM cloud's accuracy cannot be verified.

Professional Practice

In professional practice, SLAM is most powerful as part of a combined workflow rather than as a standalone answer. A common and effective pattern is to use SLAM to capture the entire site rapidly for overall context, layout and circulation, and then add terrestrial laser scanning or close-range photogrammetry only where accurate detail is required — a carved facade, a structurally significant element, a space needing a measured drawing. The SLAM cloud gives fast, complete framework; the detailed methods give accuracy exactly where it is needed. This spends the expensive, slow methods only where they earn their cost.

The professional obligation, as with any method, is to state the accuracy honestly. A SLAM survey delivered with its real accuracy (typically ±20–60mm) and its method clearly stated is a valuable, appropriately-scoped deliverable. The same survey implied to be scan-grade is a misrepresentation waiting to fail when someone measures fine detail from it. Matching method to required accuracy, and declaring what was used, is the core discipline.

Finally, SLAM's low disturbance and high speed make it particularly valuable for the many heritage situations where access is the binding constraint — active places of worship, occupied buildings, emergency records. Being able to capture a complete, adequately-accurate record of a complex space in a short, respectful visit is a genuine capability, provided it is used for the jobs it fits and not stretched to the ones it does not.