How to Choose Floor-Plan Cut Heights from a Point Cloud: A Reproducible Multi-Slice Method

Choose a point cloud floor plan slice height by reference to the plan’s intended use, the local finished-floor level, and the vertical position of the features that must appear. Use a thin primary band rather than a zero-thickness plane, then compare secondary bands above and below it. Record the height, thickness, exceptions, projection rules, and QA overlay. A fixed 1.0 m or 1.2 m value can be an example in one project, but it is not a reproducible method for every building.

This distinction matters because a floor plan is a controlled representation of three-dimensional evidence. Walls may be stable across most of a storey, while windows, deep niches, sloping faces, closed doors, counters, and low obstructions appear differently as the horizontal cut moves. The drafting method must state which geometry is actually cut, which is projected from another height, and which remains uncertain.

A single arbitrary height is not a method

A number without a purpose does not explain why the chosen band represents the requested plan. The same height can pass cleanly through one room and intersect shelving, decorative mouldings, sloped walls, or window spandrels in another. It may show a doorway in one partition but meet a closed leaf or glazed panel elsewhere. A technician who repeats only the number may therefore obtain a different-looking plan from the same storey.

Research supports treating height as a consequential method parameter. Zhao et al. explain that different indoor components appear in slices taken at different heights and compare those slices to extract doors and beams. In a purpose-specific evacuation study, D’Emilia et al. examined multiple horizontal layers and found that the operating procedure and changing section height affected the variability of derived passage geometry. Their numerical results belong to their cathedral test case, not to a universal CAD tolerance. The transferable lesson is that the measurand and the slicing procedure must be defined together.

A reproducible method answers five questions: What is the vertical reference? What must the plan support? Which band defines cut geometry? Which other bands resolve features that the primary band misses? How are conflicts and unsupported areas disclosed?

Establish the local vertical reference and the plan’s use

Begin with the registered point cloud supplied by the client, surveyor, or reality-capture provider. Confirm units, orientation, storey limits, and the project coordinate framework before deciding a cut. This is downstream source review, not a replacement for registration or survey control. If those upstream conditions are unclear, resolve them before drafting. The separate guide to preparing point-cloud data for processing covers that handoff boundary.

Define the cut height as an offset from an identified local floor datum, not merely as a global Z coordinate. A storey can contain steps, ramps, raised platforms, sunken rooms, sloping floors, or separate wings tied to different local levels. One global elevation may therefore sit at different functional heights above the floor. Create local zones when the floor reference changes materially and record how their plan geometry is reconciled.

Use a primary band, not a zero-thickness plane

A point cloud contains discrete samples. An infinitely thin mathematical plane may intersect few or no points, while a very thick slice mixes geometry from different heights. Define the primary band by a centre height hp above the local datum and a thickness t. Its vertical interval is [hp - t/2, hp + t/2].

Band thickness is a viewing and evidence-selection parameter. It is not the scanner accuracy, registration error, drawing tolerance, or line width. Select it by testing whether the band provides enough point support for continuous wall intersections without blending together returns from sloped faces, trim, furniture, adjacent levels, or several depths of a recess. If continuity is poor, first distinguish sparse coverage from an overly narrow band. Increasing thickness cannot recover a surface that was never observed.

Compare candidate and secondary slices

1. Identify candidate bands that intersect the largest continuous set of wall and column surfaces required by the scope.
2. Reject candidates dominated by transient clutter, dense furniture, mouldings, railings, or surfaces that confuse the wall body.
3. Compare line position across nearby bands. A stable wall face should remain within the project’s agreed representation rule. A moving outline signals slope, curvature, damage, ornament, or mixed depths that require an explicit decision.
4. Select the primary band that best supports the dominant plan purpose, then create targeted secondary bands for unresolved elements.
5. Save the comparison view or record its parameters so another technician can repeat the review.

Each secondary band should answer a named question. It can test a low obstruction, locate a window opening, or reveal a beam or wall return. Babacan et al. combine geometric primitives with object evidence, reminding us that a line’s meaning cannot always be decided from horizontal position alone.

Resolve cut, projected, symbolic, and inferred geometry

The final drawing should not pretend that every visible line came from the primary band. Use a provenance rule that survives into layers, line types, or the method record. A compact code can distinguish C for geometry intersected by the primary cut, S for geometry taken from a documented secondary band, P↑ and P↓ for geometry projected from above or below, and I for an inference permitted by scope and clearly disclosed. Unobserved geometry should remain a gap or limitation, not receive an inferred line by default.

Conventional symbols need the same care. A door swing arc communicates operation; it is not a measured arc. A window symbol may combine measured jamb positions with the project convention. Supplied images can assist interpretation, but they cannot create measurable geometry where the cloud has no support.

Adapt the method to the plan’s intended use

A circulation plan illustrates why purpose comes first. The D’Emilia study included obstacles across a defined vertical range because its measurand concerned evacuation passage geometry. The responsible specialist must define such an envelope before the drawing method can represent it. ENGINYRING supplies downstream measured geometry; it does not design an egress strategy or certify compliance.

For area measurement, the cut is subordinate to the named standard and its boundary rules. A simplified background may use a representative stable wall line. In both cases, state the rule: orthogonalization is a representation decision, not an automatic correction.

Record the method so another technician can reproduce it

A concise record could state: “Primary band PB-01 is referenced to F-01. Window W-14 comes from secondary band SB-03 and uses the projected-opening convention. The low recess at C-02 comes from SB-01. The unobserved return at R-08 remains open under limitation L-02.” The codes are project-specific; the logic is repeatable.

Verify the drawing with a controlled overlay

Overlay the CAD linework on the saved primary band, then inspect each documented secondary view without changing the coordinate framework. Review the evidence behind the linework instead of judging graphic neatness alone.

• Confirm that each cut-weight wall, column, and partition is supported by the primary band or an explicit override.
• Keep projected features on the agreed layer and line type.
• Review openings at their secondary height so a leaf, curtain, glass return, or clutter is not treated as wall.
• At irregular walls, record whether the drawing preserves a declared outline, averages a range, or adds a variation note.
• Check local datums at steps, ramps, and split levels.
• Leave unobserved geometry open, qualified, or excluded. A clean closure is not evidence.
• Freeze the accepted band parameters and revision with the plan.

Worked scope example for a split-level office

Consider a hypothetical office with a raised meeting suite, glazed partitions, high-sill windows, a low corridor recess, and façade storage. One global Z slice crosses the main office at a useful height but sits lower relative to the raised suite, passes through storage, and stays below the windows.

The solution creates local datums F-01 and F-02 and compares candidate bands at equivalent offsets. PB-01 gives continuous wall evidence without merging storage depth into the façade; PB-02 applies the same criterion in the raised suite. A local secondary band locates the windows, while a lower band records the circulation recess. One weak glazed return remains qualified because no band supports a stable edge.

A reviewer can reconstruct both primary bands, inspect the targeted secondary evidence, and see why one return was not closed. That makes the method suitable for a specification, thesis, or controlled production workflow.

Keep capture, drafting, and design responsibilities separate

The client’s chosen surveyor or capture provider remains responsible for field acquisition, registration, and the agreed control framework. The drafting party reviews the supplied dataset, selects the sectioning method for the requested representation, and identifies limitations. The architect, surveyor, fire specialist, area-measurement professional, or other responsible consultant defines any use-specific rule and decides whether the plan is suitable for its regulated or design purpose.

The choice between a 2D plan and a 3D geometry deliverable should also be made before the method is fixed. The comparison of Scan-to-CAD and 3D geometry outputs explains that broader decision. Once a floor plan is selected, its cut method should be treated as part of the deliverable specification, not as an undocumented software setting.

If your client or surveyor has already supplied a registered point cloud, use ENGINYRING’s point cloud to CAD service to discuss a project-specific floor-plan scope, including the required sectioning record, representation rules, and source-data limitations.