Can LiDAR See Through Walls?

Every project kickoff call surfaces the same question: “Did the scanner pick up the pipes and conduit inside the walls?” No - and the physics behind that limit is worth understanding precisely, because it determines how you scope the work, what supplemental tools you need, and where real project risk lives.

---

The Short Answer: No - and Here Is the Physics Behind That

LiDAR (Light Detection and Ranging) fires laser pulses in the near-infrared spectrum - 905 nm for time-of-flight systems, or ~1550 nm for some long-range units. Those photons travel at the speed of light, hit a solid surface, and bounce back to the sensor. The instrument measures either the round-trip time (time-of-flight) or the phase shift of a continuous beam to calculate distance. The operative word is optical, not acoustic, not electromagnetic microwave. Photons at near-IR wavelengths do not transmit through opaque materials. A single layer of 5/8-inch gypsum board stops them completely.

The confusion usually comes from the word “radar” embedded in “LiDAR.” Ground-penetrating radar (GPR) uses microwave frequencies in the 100 MHz-2.6 GHz range - wavelengths measured in centimeters to meters - which penetrate non-metallic materials and reflect off rebar, conduit, and pipe inside a slab or wall. LiDAR uses wavelengths measured in nanometers. They are categorically different instruments solving different problems.

The Trimble X7 official datasheet specifies a range accuracy of 2 mm and a 3D point accuracy of approximately 4 mm at 10 m. Scanners such as the Leica RTC360 deliver 1.9 mm at 10 m. Neither can read through a sheet of drywall, a CMU block, a pane of glass, or a concrete slab - that constraint is optical physics, not an instrument limitation that a better scanner resolves.

---

What LiDAR Does Capture: A Room-by-Room Reality Check

What phase-shift scanners capture is every visible, reflective surface in line-of-sight: floors, structural ceilings, exposed columns, beams, window reveals, door frames, millwork, casework, exposed MEP (ductwork, conduit trays, pipes, hangers), rooftop equipment, parapets, curtainwall profiles, stairwell stringers, handrails, and elevator shaft walls where access panels are open.

The accuracy that makes all of this useful:

Sub-5 mm accuracy across a full room volume simultaneously - no tape measure, total station sketch, or hand redline delivers that. A full 50,000 sq ft office floor scanned at standard density generates roughly 8-25 GB of .E57 point cloud data depending on scan spacing and overlap. That file is a time-stamped dimensional record of the building as it existed on scan day.

Glass surfaces require a specific callout. Most architectural glass partially transmits near-IR wavelengths. In practice, the scanner frequently returns two hits: a valid return off the glass surface itself, and a faint ghost return from a surface behind the glass - the floor beyond a window wall, or the back of a storefront. In a point cloud, those ghost returns appear as a thin duplicate surface offset 50-200 mm inside the actual glass plane. During point cloud cleaning in Autodesk ReCap or CloudCompare, those ghost returns must be flagged and either removed or annotated. We document glass surfaces explicitly in every deliverable scope note.

Stairwells, open shafts, atriums: LiDAR travels through open air without restriction. A scanner positioned at a stair landing captures every tread, riser, wall, and guardrail in line-of-sight. The constraint is always solid matter, never air.

For a detailed breakdown of what professional 3D laser scanning services capture and deliver, see our services page.

---

The Real Problem: Occlusion, Not Penetration

What actually drives field time and cost on most building scan projects is not wall penetration - it is occlusion: shadow zones created when any object stands between the scanner and a target surface.

A steel column creates a shadow. A modular workstation creates a shadow. A bank of lockers creates a shadow. A 36-inch supply duct running the length of a mechanical room creates a massive shadow on everything behind it. None of those shadow zones represent a failure to “see through” anything. They are geometry that was not in the scanner’s line of sight from that setup position. The solution is more scanner setups, planned in advance against the floor plan.

Setup count varies dramatically by space type:

A cluttered mechanical room with freestanding equipment, conduit bundles, cable trays, and AHUs may require 8-12 scanner positions per room to close all shadow zones. An open office bay may need 2-4.

Registration residuals and what degrades them: We use cloud-to-cloud plus target-based registration in Trimble Perspective, targeting RMS residuals below 3 mm for construction-grade work. Two conditions reliably push residuals above that threshold. First, scanner spacing beyond 10 m between consecutive setups in a featureless corridor - low surface texture gives the cloud-to-cloud algorithm insufficient geometry to lock onto, so we supplement with spherical or checkerboard targets at those stations. Second, temperature differentials greater than 10°C across a multi-day campaign - overnight thermal cycling causes physical targets to shift relative to the building structure, corrupting residuals on setups that span a temperature swing. On multi-day projects in cold-weather climates, we re-survey at least two control targets each morning before resuming capture. Any setup pair showing residuals above 5 mm triggers same-day investigation before the crew demobilizes.

Learn more about how point cloud registration works and why RMS error matters.

---

The Hidden Building Problem: What Actually Lives Inside Walls

Structural steel, rebar, post-tension cables, conduit, plumbing supply and return lines, fire suppression piping, insulation, vapor barriers, blocking - none of it is captured by any surface-based scanner, regardless of brand or price point.

The assumption that creates real project risk: “You scanned the whole building, so we have everything, right?” We have everything visible. The wall surface geometry is documented to ±2-3 mm. What is inside the wall is a separate data-collection problem, and conflating the two creates exposure on any project where walls will be opened, cores drilled, or demolition performed.

The tools that actually penetrate - with real specs:

GPR (ground-penetrating radar): The GSSI SIR-4000 operating at 1.6 GHz resolves conduit and rebar in a 6-inch CMU wall. That same 1.6 GHz antenna loses reliable resolution at increased depth through reinforced concrete - signal attenuation from rebar and moisture degrades the return below usable signal-to-noise ratio. For post-tension slab work, a 900 MHz antenna penetrates deeper but at coarser spatial resolution; note that GSSI and MALÅ (Guideline Geo) are entirely separate manufacturers, and the specific depth range achievable depends on material composition, moisture content, and rebar density. The Proceq GP8000, a newer handheld GPR unit, operates at 200 MHz-4000 MHz (0.2-4.0 GHz) swept frequency and is well-suited for wall surveys where the target depth is under 12 inches - it identifies conduit, rebar, and pipe within a standard 3.5-inch or 6-inch stud wall. None of these tools provide 3D geometry; they provide a 2D slice at a fixed antenna path, which must be co-registered with LiDAR surface geometry to be positioned in the building model.

Borescope cameras: Inserted through a pilot hole (typically 5/8-inch to 1-inch diameter, drilled through the finish surface into the wall cavity) to visually confirm cavity conditions - insulation type, pipe material, stud spacing, presence of blocking. A 9mm flexible borescope handles tight access; a 17mm rigid borescope is used where the pilot hole size is acceptable to the owner. This is a destructive operation requiring patch and paint after, but it costs $50-$150 per location versus the risk of a demo-day surprise.

Record drawings: Worth pulling from the file. After 20+ years of renovation cycles, they are frequently incomplete or contradicted by field conditions - but they identify original routing logic and structural details that inform where GPR and borescope effort should concentrate.

When hidden conditions matter most: adaptive reuse projects where exterior walls will be stripped to structure, gut renovations where MEP is being fully replaced, and historic buildings where wall construction is uncertain. In those cases, why outdated building drawings create real project risk details the downstream consequences of working from incomplete data. Best practice: combine the LiDAR scan (surface geometry, ±2-3 mm) with a targeted GPR survey (subsurface features) and record drawing review before demolition begins.

---

Above-Ceiling and Below-Floor: The Access Problem and How Pros Solve It

Drop ceilings are the single largest source of “missing data” complaints on scan-to-BIM projects. A scanner positioned below a 2x2 acoustical tile ceiling sees exactly one surface above it: the tile plane. Everything above - supply and return ductwork, structural deck, conduit bundles, sprinkler mains, data cable trays, fire alarm wiring - is completely invisible.

Standard workflow for above-ceiling capture: Before the scan day, we coordinate with the GC or facilities manager to pop ceiling tiles at representative bays - typically one access point every 400-600 sq ft of ceiling, plus at every major MEP change of direction. At each opening, we place our terrestrial scanner on a low-profile mount and run 1-2 setups above the ceiling plane. Those setups register into the same point cloud as the below-ceiling data, giving the Revit modeler a complete picture of the plenum.

Cost and time implication:

Those delta figures need to be in the RFP, not discovered after the field crew has packed up.

Raised-access floors in data centers, labs, and trading floors follow the same logic. Floor panels must be pulled at key locations. The scanner captures below-floor conduit, power distribution, and cooling infrastructure - but only if the panel is open when the scanner fires.

LOD implications are direct:

Modeling above-ceiling MEP to LOD 300 requires physical access and actual scan data. If the scope document says “LOD 300 MEP” but the plenum was never accessed, the model contains guesses, not measurements. For the full breakdown, see our post on capturing above-ceiling MEP geometry for renovation models.

---

Special Cases: Glass, Water, Vegetation, and Dark Surfaces

Real-world buildings present surfaces that behave unpredictably with near-IR laser pulses:

Dark-object sensitivity is directly relevant in server rooms and data centers where equipment is matte black or dark gray. In high-sensitivity scan mode with scanner spacing tightened to one position every 200-250 sq ft, effective range and data quality on dark surfaces improve measurably. Field logs should document which sensitivity mode was active at each setup, so QA reviewers understand the noise floor.

For facade-specific challenges including glazing surveys, see our post on how LiDAR compares to photogrammetry and terrestrial laser scanning.

---

The Workflow That Gets You Complete Building Data Despite These Limits

Our standard six-step process for a commercial building scan:

Step 1 - Pre-scan planning. Pull existing drawings (even outdated ones identify wall locations and ceiling heights). Walk the space or review photos. Mark high-occlusion zones: mechanical rooms, server closets, dense storage, above-ceiling plenum access points. Confirm with facilities whether ceiling tiles and floor panels will be pulled. Define scope boundary clearly: which floors, which plenums, which mechanical rooms, which exterior facades.

Step 2 - Field capture. We deploy the Trimble X7, which captures a full 360° scan in approximately 2 minutes per setup. We log every setup position with a field sketch or photo, noting setup number, time, and access conditions. A 50,000 sq ft floor with standard below-ceiling scope typically runs 4-6 hours of field time with a two-person crew. A 50,000 sq ft floor at standard density generates 8-25 GB of .E57 data depending on scan spacing and overlap - store that file permanently. That E57 is a timestamped existing-conditions record: when a wall gets demolished in a future renovation, when a dispute arises about as-built conditions, the cloud is the documentation. Remobilization to rescan a single floor typically costs $4,000-$8,000; storing 25 GB of .E57 indefinitely costs less than $2/month in cloud storage.

Step 3 - Registration. Import all setups into Trimble Perspective (our primary registration platform). Run cloud-to-cloud registration for rough alignment, then refine with target-based registration where spherical or checkerboard targets were placed. Target RMS residual: < 3 mm for construction-grade as-built documentation. Residuals degrade past that threshold in two predictable situations: scanner spacing beyond 10 m in low-feature corridors (supplement with physical targets), and temperature differentials above 10°C across a multi-day campaign (re-survey control targets each morning). Any setup pair above 5 mm RMS triggers investigation before demobilization.

Step 4 - QA review. Load the registered cloud in Autodesk ReCap. Fly through every room, looking for shadow zones, missing surfaces, and anomalous returns (ghost glass reflections, water returns). If critical gaps exist, same-day return to site is always preferable to a second mobilization.

Step 5 - Delivery. Output formats: .RCP/.RCS for direct Revit import, .E57 for interoperability across platforms (CloudCompare, FARO Scene, Bentley), .LAS/.LAZ for GIS workflows. Every delivery package includes a scope-of-capture document - one page stating what was scanned, what was not, registration accuracy achieved, and any known data gaps. Standard turnaround for a registered cloud delivery on a single floor is 5-7 business days from scan day. See what a laser scanning deliverable package actually contains for the full breakdown.

Step 6 - Modeling. The Revit modeler works directly from the point cloud, snapping model elements to measured geometry. Walls, floors, columns, beams, ductwork, and pipe all go in at measured locations. Anything not in the point cloud - hidden conduit, above-ceiling MEP that was never accessed - does not get modeled without supplemental data. The model is only as complete as the scan.

For the full modeling workflow, see our scan-to-BIM workflow from registered cloud to Revit model and the companion point cloud to Revit workflow guide.

---

Emerging Tech: Is “Seeing Through Walls” Ever Possible with Scanning?

The commercially realistic picture in 2025-2026:

Terahertz (THz) imaging operates between microwave and infrared (0.1-10 THz) and can image through thin non-metallic materials in controlled conditions. Research has demonstrated promising spatial resolution through thin plastic and composite panels, but the physics degrades quickly in building envelope contexts: THz signal penetrating concrete loses usable signal-to-noise ratio rapidly because concrete’s free-water content absorbs THz energy aggressively. Any metallic content - rebar, wire mesh, foil-faced insulation - blocks THz transmission entirely, the same way it blocks visible light. Field-deployable units for AEC building applications do not exist as of 2025. Lab systems require controlled humidity, short standoff distances (typically under 30 cm), and scan times measured in minutes per square foot. Not commercially mainstream.

GPR + LiDAR data fusion is an active research area. The core concept - registering GPR subsurface datasets with LiDAR surface geometry into a unified BIM coordinate system - is technically feasible for slab-on-grade and pavement applications where the GPR antenna travels a defined 2D grid. For vertical construction in a commercial building, the geometry of fusing a hand-pushed cart trace along a wall face with a 3D point cloud introduces position error from cart wheel slip and antenna liftoff. Research groups have published frameworks for this, with co-registration accuracy for subsurface features referenced to LiDAR surface control that is adequate for “avoid this zone before drilling” but generally insufficient for precise MEP routing in a BIM model. The specific accuracy achievable varies by site conditions, antenna frequency, and registration methodology.

“AI wall fill” - correcting the record: Some marketing language implies that Autodesk ReCap or Leica Cyclone REGISTER 360 can generate geometry in occluded zones. Neither product has a generative wall-fill interpolation feature as of 2025. What ReCap and Cyclone REGISTER 360 do offer is noise filtering and visualization tools that make data gaps visible - but they do not invent geometry that was not scanned. The tool that does offer AI-based surface infill is Matterport’s AI Infill for its own capture format, and Cyclone 3DR (a separate Leica product distinct from Cyclone REGISTER 360) offers surface reconstruction and mesh-fill tools that fit a continuous surface through sparse point data - useful for smoothing a concrete slab return, not for fabricating hidden MEP. We flag any mesh-reconstructed geometry explicitly in every deliverable.

Indoor drone LiDAR (DJI Matrice 300 with Zenmuse L2, Flyability Elios 3 with integrated LiDAR payload) gets the scanner into positions a tripod cannot reach - 40-foot atrium ceilings, boiler room tops, confined mechanical spaces. The physics are identical: the drone-mounted scanner still captures only visible surfaces. The key tradeoff is accuracy: per DJI’s official Zenmuse L2 specifications, the system delivers 5 cm horizontal and 4 cm vertical accuracy at 150 m using RTK/IMU positioning - roughly 10-20x worse than a static terrestrial scanner at the same range. For LOD 200 documentation of hard-to-access voids that is acceptable; for LOD 300 or 350 work requiring ±5-10 mm accuracy, the drone data is supplemental, not primary. Drone platforms such as the Flyability Elios 3 are used by operators for confined-space industrial work (tanks, ductwork interiors, ceiling voids above sealed plenums), with output registered to the primary static-scanner cloud as a reference layer.

For the foreseeable future, subsurface data requires a subsurface instrument. LiDAR surface data is exceptionally accurate - and it covers everything it can see. The limit is physics, and knowing exactly where that limit sits is what makes a properly scoped project succeed.

---

Practical Takeaways for Architects, Engineers, and Owners

Scope your scan in the RFP, not after. Every surface that must be captured - above-ceiling, below-floor, mechanical rooms, exterior facades, roof - needs to be called out explicitly before the crew mobilizes. “Get everything” is not a scope. It is a change-order waiting to happen.

Budget for access and supplemental tools separately:

For what a laser scanning deliverable specification should include, see our template resource.

Use scan data for what it does well: exterior dimensions to ±2-3 mm, floor-to-floor heights, column grid, room areas, facade profiles, window and door opening geometry, as-built MEP routing where exposed, existing-conditions BIM for renovation design at LOD 300. A registered point cloud from our Trimble X7 delivers those results consistently and repeatably.

Include a scope-of-capture summary in every deliverable package. One page. What was scanned, what was not, accuracy achieved, known gaps. Downstream users - structural engineers, MEP designers, GCs reading the model months later - need to know whether that above-ceiling plenum was actually captured or left as a data gap.

Store the raw .E57 cloud permanently. A full 50,000 sq ft floor at standard scan density generates 8-25 GB of .E57 data. At current cloud storage rates, that costs under $2/month to retain indefinitely. Remobilizing a two-person crew for a rescan of a single floor runs $4,000-$8,000 before travel. The math is obvious.

---

FAQ

Can LiDAR detect pipes or wires inside walls?
No. LiDAR reflects off surfaces only - near-IR photons do not transmit through drywall, concrete, or any other opaque material. For pipes, conduit, and rebar inside walls or slabs, ground-penetrating radar is the correct tool. A 1.6 GHz GPR antenna (e.g., GSSI SIR-4000) can resolve conduit in a 6-inch CMU wall, with resolution degrading at increased depth through reinforced concrete due to signal attenuation from rebar and moisture. Many projects combine a LiDAR scan for surface geometry with a GPR sweep for subsurface features, then co-register both datasets into a common BIM coordinate system.

Does LiDAR work through glass?
Partially and unpredictably. Most architectural glass partially reflects and partially transmits near-IR wavelengths. You may get a valid return from the glass surface, a ghost return from a surface behind it, or no return at all when the beam strikes at a low angle. Best practice: scan from multiple angles, flag glass surfaces explicitly in the point cloud deliverable, and remove or annotate ghost returns during QA cleaning in ReCap or CloudCompare.

Why does my point cloud have dark gaps (shadow zones)?
Those are occlusion shadows, not wall-penetration failures. Any object between the scanner and a target surface blocks the laser - a desk, a duct, a column, a piece of equipment. The solution is additional scanner setups positioned to surround obstructions. A well-planned crew maps setup positions against the floor plan before scan day, targeting 15-25 setups per 10,000 sq ft in a moderately furnished commercial space, scaling up to 40-60 per 10,000 sq ft in dense mechanical rooms.

Can a drone LiDAR scan the inside of a building?
Yes, and it is increasingly useful for access-constrained spaces. The DJI Matrice with Zenmuse L2 and the Flyability Elios 3 get the scanner into atria, confined mechanical spaces, and tall voids a tripod cannot reach. The accuracy tradeoff is significant: per DJI’s official specifications, the Zenmuse L2 delivers 5 cm horizontal and 4 cm vertical accuracy at 150 m using RTK/IMU positioning - roughly 10-20x worse than a static terrestrial scanner. That is acceptable for LOD 200 documentation of hard-to-reach voids; it is not sufficient for LOD 300 or 350 work. We use drone LiDAR as a supplemental layer registered to the primary static-scanner cloud, not as a replacement for it.

What accuracy does a LiDAR building scan deliver?
The Trimble X7 datasheet specifies a range accuracy of 2 mm and a 3D point accuracy of approximately 4 mm at 10 m. The Leica RTC360 delivers 1.9 mm single-point accuracy at 10 m. After cloud-to-cloud registration across all scan setups, a well-controlled project holds ±2-3 mm globally - well within the tolerance for construction documentation, renovation design, and scan-to-BIM at LOD 300. Residuals degrade in low-feature corridors with scanner spacing beyond 10 m, or when temperature differentials above 10°C occur across a multi-day campaign.

Is above-ceiling MEP captured in a standard building scan?
Not automatically. A scanner placed below a drop ceiling sees only the tile plane. Capturing above-ceiling MEP requires ceiling tiles opened at representative bays - typically one access point every 400-600 sq ft - with 1-2 scanner setups above the ceiling at each location. This scope must be specified in the project RFP. It typically adds 15-30% to total field time and flows directly into project cost. If above-ceiling scope is not in writing, assume it was not captured.

How large are .E57 point cloud files, and do we need to keep them?
A single floor scan of 50,000 sq ft at standard density typically generates 8-25 GB of .E57 data, depending on scan spacing and overlap settings. Yes, keep them permanently. That file is a timestamped existing-conditions record of the building as it existed on scan day - usable for future renovations, dispute documentation, and insurance claims. At current cloud storage rates, retaining 25 GB costs under $2/month. A rescan mobilization costs $4,000-$8,000 minimum.

---

Ready to Get Complete Building Data - With No Surprises?

We deploy the Trimble X7 terrestrial laser scanner, along with handheld scanners for detailed and parts work, on commercial, institutional, and industrial projects; based in the New York metro area, we travel nationwide to your site. Standard turnaround is 5-7 business days for a registered point cloud delivery on a single floor - delivered in .RCP, .E57, and .LAS formats with a one-page scope-of-capture summary documenting registration accuracy, known data gaps, and access conditions. For projects requiring Revit models at LOD 200, 300, or 350, we carry that through from the registered cloud.

Every engagement starts with a scoping call where we review your floor plans, confirm access conditions, and define exactly what will and will not be in the deliverable - in writing, before mobilization.

Contact us to discuss your project scope, timeline, and budget.