Registered vs Unregistered Point Clouds

A registered point cloud is scan data that has been aligned into one coordinate space, so every scan position reads as a single measured environment. An unregistered point cloud is the opposite: separate scan stations or export files that have not been tied together yet, and that will not line up well enough to model, draft, or measure against. If you are about to start a Revit model or a CAD drawing set, this is the difference between data you can build on and data that still needs work.

The reason it matters is that “we have scan files” and “we have a usable point cloud” are not the same statement. A scan provider can hand off a folder, a USB drive, or an E57 file that looks complete, and a buyer reasonably assumes it is ready for production. Sometimes it is. Often it is the raw material, and the alignment step that makes it trustworthy has not been done or has not been verified. Confirm registration before anyone traces a wall or places a level.

What registration actually does

A survey-grade terrestrial scanner captures from a fixed position, then moves to the next. Each station sees only part of the building: one scan covers a corridor, the next covers a room off that corridor, a third covers the stairwell. Each of those scans has its own internal origin. Registration is the process of solving how those stations relate to each other in space, so the corridor, room, and stairwell share one consistent set of coordinates.

There are two common ways to do it. Target-based registration uses physical references (spheres or checkerboard targets) placed in overlapping scan coverage, which the software matches between stations. Cloud-to-cloud registration aligns the overlapping geometry itself, no targets required, which is how many modern scanners self-register in the field using onboard sensors and surface matching. Most real projects use a mix, then a person reviews the result. Good registration is measured, not assumed: a registration report lists how many stations were tied, the overlap between them, and the residual error, often in the low single-digit millimeters for interior work.

When that solve is weak or skipped, the symptoms are predictable. Walls appear twice, a few millimeters or a few inches apart. Corners look fuzzy because two stations disagree about where the corner is. A stair does not connect cleanly between floors. The data can still look impressive on screen while quietly hiding the fact that no single coordinate is reliable.

Registered vs unregistered, side by side

The right side of that table is not a failure. It is simply an earlier stage of the same data. The problem only appears when an earlier stage gets treated as a finished deliverable.

File format does not equal registered

A common point of confusion is whether an E57 file is registered. The format alone does not answer that. E57 is an open exchange format for point cloud and imagery data; it is portable and widely supported, but a single E57 can hold one merged registered cloud, or it can hold a stack of individual stations that were exported without ever being aligned. The extension tells you nothing about the solve.

RCP and RCS are Autodesk’s point cloud formats for Revit and AutoCAD. An RCP is a project file that references one or more RCS scan files; it exists because Autodesk software needs the cloud indexed into its own structure before you can attach it. Having an RCP means the data was processed for Autodesk, but it still does not guarantee the underlying registration was good. What tells you the truth is the registration report and a quick look at the cloud in a viewer, not the filename. For more on the Revit side specifically, see how to import a point cloud into Revit and our point cloud to Revit workflow.

How registration shows up in the deliverable

For a Revit model, registration is what holds wall alignment, floor-to-floor relationships, opening positions, stair geometry, roof planes, and MEP routing in agreement. Build on a weak registration and those relationships inherit the error.

For CAD drawings, registration sits underneath every line. Picture a long corridor captured from two stations that disagree by half an inch. A drafter traces the wall from one station at the near end and from the other station at the far end. The resulting plan looks clean and professional while encoding a wall that is not straight in reality. Nobody catches it until a contractor measures on site. Solid registration removes that ambiguity before the drafting starts, which is also why we treat it as a checkpoint in our scan-to-BIM workflow rather than an afterthought.

When unregistered data is exactly what you want

Unregistered files are not useless, and you should not throw them away. If you are commissioning the registration, the raw scanner project is precisely what the team needs: it carries the original station data, intensity, and sometimes color and sensor metadata that a flattened export drops. Raw files are also what lets a team recover a missing export, reprocess a project, split a cloud by floor, convert to a different format, or troubleshoot a deliverable that someone else produced. The single rule is to not treat raw or partial data as a finished product until the alignment has been reviewed.

How to tell what you actually have

If you have scan data and are not sure where it stands, the fastest path is to send it for review rather than guess. Send the full scanner project folder intact (renaming or splitting files often breaks the folder relationships the project depends on), any RCP/RCS or E57 exports, the registration report if one exists, the building address or a short description, your target deliverable, and the software your team plans to use. From there it takes a short look to say whether the data is registered, whether coverage is complete, and what, if anything, it still needs.

If you are scoping new work instead, the questions worth asking a provider are simple: Is the data registered, and what is the residual error? Has registration QA been done, and can you share the report? Are all stations included and all floors tied together? Are interior and exterior scans aligned to each other? Which formats do I receive, including the Autodesk-ready version? Is there a defined coordinate system? Those answers separate a usable dataset from a folder of files in about two minutes.

If you have a point cloud and you are not sure it is ready for production, send it to us for a registration check before your modeling or drafting team starts. A short review up front is far cheaper than rework once a model is half built.

FAQ

What is point cloud registration?

It is the process of aligning the separate scan positions captured during a survey into one shared coordinate space, so the whole building reads as a single measured dataset. It can be done with physical targets, by matching overlapping geometry (cloud-to-cloud), or both, and the result is checked against a measured residual error.

How does 3D laser scanning work?

A survey-grade terrestrial scanner sits at a fixed position and sweeps a laser across everything it can see, recording millions of distance measurements per scan. It moves to new positions until coverage is complete, often a full day on site for a building. Those individual scans are then registered together into the point cloud you receive.

Is my E57 file registered?

You cannot tell from the format. An E57 can contain one merged, registered cloud or a set of unaligned stations. The way to confirm is to check the registration report and open the file in a viewer to see whether the geometry is consistent, not to rely on the extension.

What is scan to BIM?

Scan-to-BIM is the process of building an intelligent 3D model (typically in Revit, at a defined level of detail such as LOD 200 or LOD 300) from a registered point cloud. It depends on registration first: the model can only be as accurate as the aligned data underneath it. Learn more on our scan-to-BIM service page.

Can I get CAD drawings straight from an unregistered point cloud?

Not reliably. Point cloud to CAD conversion traces linework directly from the cloud, so any disagreement between unaligned stations is baked into the plan, section, or elevation. Register and QA the data first, then draft.

A registered point cloud is a project dataset you can trust. An unregistered point cloud is scan data that still owes you an alignment step. If your next move is a Revit model, a DWG set, as-built documentation, or a restoration record, confirm registration before production starts. Everything downstream depends on the scan data lining up first. You can also use our 3D laser scanning service (registration included) if the data needs alignment before that work begins.