What Point Cloud Processing Services Deliver

Point cloud processing turns collected observations into an organised dataset that other teams can use. The service may include registration, cleaning, reference alignment, segmentation and export, but those tasks should be agreed explicitly. Processing a cloud is also different from producing a measured drawing or a building model from it.

By GeobiroFirst published Updated 4 min read
AI illustration of a tripod laser scanner capturing a building interior as a point cloud.
AI-generated editorial illustration. It does not depict a documented Geobiro project.

Begin with an input review

Before production, check which files and supporting records are available. A processor may need source scans, existing registration information, imagery, control records and field notes. Identify the capture dates, equipment formats and any previous transformations or edits.

Ask for an initial assessment that lists missing inputs and likely limitations. If the material comes from several suppliers or visits, clarify whether it shares a common reference. Do not assume that similarly named files represent the same coordinate system or stage of the site. Resolve these issues before substantial processing effort is committed.

Agree the assembly and reference work

Registration is required where separate scans have not yet been reliably assembled. The scope should explain whether the provider is creating that assembly, reviewing an existing one or only converting an already accepted dataset. Autodesk's ReCap learning material illustrates registration review before indexing and use of the project.

Reference alignment should also be explicit. State the intended units and coordinate framework, and supply the relevant survey information. If an existing registration cannot be supported by adequate evidence, ask for options and limitations. A conversion service should not silently imply that the underlying survey has been verified.

Define what cleaning and organisation mean

Cleaning should have a purpose. The client might want transient people separated, distant context cropped or obvious stray observations removed. Agree the treatment of temporary equipment and objects that could matter to later interpretation. Preserve the source record before making changes.

Organisation may include zones, floors or other project divisions. Request names that make sense to the receiving team and document boundaries. If point reduction is proposed to improve usability, review a representative result to ensure relevant small features remain available. Keep a fuller master record when a reduced version is being issued.

Specify each output as a separate deliverable

A registered cloud, a mesh, CAD drawings and a BIM model serve different purposes and involve different decisions. List each required output rather than expecting it to be included under processing. Define whether attributes, surface interpretation or object identification are needed.

For exports, specify recipient applications and the information that must survive transfer, including imagery or scan structure where required. Autodesk's E57 documentation shows that export options can carry different content. Test a sample and inspect the final files, because an export completing successfully does not establish that the recipient has everything needed.

Expect a traceable handover

Request a short processing record covering the inputs, reference changes, cleaning approach and exported versions. Include relevant registration evidence, coverage notes and unresolved issues. The record should make it possible to distinguish the checked master from later working copies.

Processing cannot recreate observations that were never collected. If a required area is missing or unsuitable, document that limitation and identify the next action. The final acceptance check should confirm that the recipient can open the files, navigate the agreed zones and perform the planned work without relying on undocumented production settings.

Common questions

Can processing repair missing survey coverage?

It can organise existing evidence and identify gaps, but it cannot create a measured record of unseen surfaces. Any interpolated or reconstructed geometry must be labelled, and important gaps may require additional capture.

Does a clean point cloud include an as-built model?

Usually these are different deliverables. A model requires decisions about elements, geometry, attributes and uncertainty. Include modelling explicitly in the brief and agree the review criteria for it.

Further reading

This article was rewritten in October 2026 around the subject of the original Geobiro post. The original publication date is retained; the update date identifies the current text.

Let’s scope the work.

Tell us about your site, the information you already have and the deliverables you need. Our team can help define the next step.

Tell us what's next.

Your inquiry is sent to Geobiro. We use your details to respond to your request.