10 Point Cloud QA Checks to Run Before Scan to BIM

# 10 Point Cloud QA Checks to Run Before Scan to BIM

TL;DR

• Verify registration reports and survey control to confirm foundational point cloud accuracy.
• Check for vertical drift and correct project coordinates before modeling begins.
• Sweep the dataset for data gaps (occlusions) and unacceptable noise or artifacts.
• Confirm data density is sufficient to model features at the furthest scan ranges.
• Ensure file formats, indexing, and panoramic imagery meet project specifications.

Jump to:

How do you review registration quality? · How do you validate against survey control? · How do you check for vertical alignment drift? · Are the coordinate system and project units correct? · Is data coverage sufficient and are occlusions acceptable? · What level of noise or artifacts is acceptable? · Is the point colorization or intensity data useful? · Is the point density adequate in large, open spaces? · Are the file formats correct and indexed for performance? · Are scan positions and panoramic views complete?

This guide is for architects, BIM managers, and engineers who receive point cloud deliverables from a 3D laser scanning provider. It outlines a systematic quality assurance (QA) checklist to execute before any Scan to BIM modeling begins. Adopting this inspection process protects your project from the downstream consequences of inaccurate as-built documentation.

The purpose of these checks is to formally accept or reject the point cloud dataset as a valid foundation for your as-built model. Discovering a registration error, incorrect coordinate system, or significant data gap after modeling has started can lead to extensive rework, budget overruns, and schedule delays. By verifying the data upfront, you ensure the deliverable meets the technical requirements of the scope of work and is fit for purpose.

1. How do you review registration quality?

The registration process digitally "stitches" individual scans into a single, cohesive point cloud. Errors in this step compromise the accuracy of the entire dataset.

• How to Perform It: Locate and open the registration report, which is typically a .txt or .pdf file included with the deliverables. This report contains the statistical results of the point cloud registration. Look for key metrics like mean error (or Root Mean Square Error - RMSE), maximum error, and point overlap percentage between adjacent scans.
• Pass/Fail Criteria: Pass if the mean registration error is within the project tolerance, which for most architectural projects is below ±5mm. Pay close attention to any maximum error outliers. A few high-tension links might be acceptable in non-critical areas, but systemic high error points to a flawed registration.
• What to Do If It Fails: If the report shows errors exceeding the specified tolerance, immediately contact the scanning provider. Send them the report and ask for an explanation or reprocessing. A flawed point cloud registration invalidates all subsequent measurements.

2. How do you validate against survey control?

For projects requiring high global accuracy, the point cloud must be tied to a survey control network. This check verifies that the alignment was done correctly.

• How to Perform It: Import the point cloud (e.g., an .RCP file) into an application that can read coordinates, like Autodesk Recap, Revit, or Civil 3D. Also, import or open the survey control file (.txt, .csv) provided by a licensed surveyor. Locate the physical control targets in the point cloud and use measurement tools to query their X, Y, and Z coordinates. Compare these values to the "true" coordinates in the survey file.
• Pass/Fail Criteria: Pass if the deviation between the point cloud's control point coordinates and the survey file's coordinates is within the project's global accuracy tolerance (e.g., ±1/4" or 6mm).
• What to Do If It Fails: A failure here is critical. It means the entire dataset is translated or rotated incorrectly relative to the project datum. All modeling must stop. The provider must re-register the entire project using the correct control points and re-export the data.

3. How do you check for vertical alignment drift?

In multi-story buildings, small registration errors can accumulate, causing upper floors to "drift" or "sag" relative to lower floors.

• How to Perform It: Open the point cloud and create a full-building elevation or cross-section view. Use the measurement tool to check the floor-to-floor or floor-to-ceiling height at one end of the building. Now, go to the opposite end of the building and take the same measurement. Repeat this for several locations across the building's footprint.
• Pass/Fail Criteria: Pass if the vertical dimensions are consistent throughout the building (within tolerance). Floors should appear level and parallel unless the structure is intentionally sloped. There should be no discernible sag in long hallways or open floors.
• What to Do If It Fails: Vertical drift can severely impact the modeling of stairs, elevator shafts, and MEP risers. If you detect it, notify the provider. Correcting this often requires re-processing the registration with more robust vertical constraints between floors, a common practice when scanning large facilities like an 80,000–120,000 sq ft distribution center.

4. Are the coordinate system and project units correct?

Modeling cannot begin if the project is in the wrong coordinate system or units. This simple check prevents fundamental setup errors.

• How to Perform It: In your viewing software, inspect the project's properties or settings to find the defined units (e.g., Feet, US Survey Feet, Meters) and coordinate system (e.g., a state plane system or local origin). As a quick practical test, measure a known object like a standard 3-foot doorway.
• Pass/Fail Criteria: Pass if the units and coordinate system match what was specified in the 3D laser scanning scope of work. The test measurement should yield a logical value (e.g., the doorway measures ~3 ft, not 0.9 ft or 914 ft).
• What to Do If It Fails: This is a non-negotiable failure. The provider must re-export the entire dataset with the correct project settings. No modeling or data cleaning should be attempted until this is fixed.

5. Is data coverage sufficient and are occlusions acceptable?

Occlusions are data "shadows" or gaps caused by objects blocking the scanner's line of sight. This check ensures no critical information is missing.

• How to Perform It: Conduct a thorough visual "fly-through" of the point cloud. Navigate into every room, plenum space, and area that was part of the scanning scope. Actively look for holes in the data behind columns, large furniture, open doors, and dense equipment.
• Pass/Fail Criteria: Pass if all required elements for the specified LOD (Level of Development) are captured. Minor occlusions are unavoidable and acceptable (e.g., a small gap behind a trash can). Unacceptable failures include missing data on structural columns, main MEP racks, or entire walls hidden by temporary obstructions.
• What to Do If It Fails: If critical data is missing, document the occlusions with screenshots and send them to the scanning provider. This may necessitate a costly return trip to the site for fill-in scanning. This check underscores the importance of proper site preparation before scanning begins.

6. What level of noise or artifacts is acceptable?

Noise is the "fuzziness" of a scanned surface, while artifacts include ghosted images of moving objects. Both can complicate modeling.

• How to Perform It: Zoom closely into a flat, uniform surface like a drywall partition or concrete floor. Assess the "thickness" of the point cloud surface. Then, scan through populated areas, circulation paths, and areas near windows to look for the faint, transparent forms of people, vehicles, or equipment that were moving during the scan. Check highly reflective surfaces like mirrors or polished stainless steel for smeared or misplaced points.
• Pass/Fail Criteria: Pass if surfaces are generally "tight" and well-defined, and artifacts are minimal and isolated. A small amount of noise is inherent to all scanners. The data fails if noise is so high that it's difficult to determine the true surface location, or if ghosting artifacts obscure important existing conditions.
• What to Do If It Fails: High noise may require the provider to re-process the raw data with different filtering parameters. Significant artifacts may need to be manually cleaned from the point cloud before modeling, which could be a change order if not specified in the SOW.

7. Is the point colorization or intensity data useful?

3D scanners capture either RGB color from built-in cameras or an intensity value based on surface reflectivity. This data provides crucial context for modelers.

• How to Perform It: In your point cloud software, toggle the display setting between "RGB" and "Intensity" (sometimes called "Grayscale"). With RGB color enabled, look for areas that are completely white (overexposed) or black (underexposed), or where motion has blurred the photo texture. With intensity enabled, check that there is good contrast between different materials.
• Pass/Fail Criteria: Pass if the color or intensity data is clear enough to help a modeler distinguish between a pipe, a conduit, and a duct. The goal is context. The data fails if poor lighting, bad exposures, or low-contrast intensity makes it impossible to visually interpret the scene.
• What to Do If It Fails: Poor colorization is rarely a reason to reject a dataset on its own, as the geometry may still be accurate. However, it significantly slows down the modeling team by forcing them to guess what objects are. It should be noted in the QA report as a factor that will impact the Scan to BIM pricing and timeline.

8. Is the point density adequate in large, open spaces?

Point density decreases with distance from the scanner. In large facilities, this can result in insufficient data on far-away surfaces.

• How to Perform It: Navigate to the largest open spaces in the project, such as an atrium, warehouse, or long corridor. Position your view near one wall and look across the space to the farthest opposite wall. Zoom in and inspect the density of points on that far surface.
• Pass/Fail Criteria: Pass if the point spacing on the farthest surfaces is still tight enough to define the features required by the modeling SOW. For example, if you need to model window mullions from 150 feet away, you need to see more than just a few sparse points on the window.
• What to Do If It Fails: Insufficient density at range is a result of a poor scan-planning strategy. It prevents accurate modeling of distant objects. The only solution is for the provider to return to the site and capture additional scans from closer positions, which has significant cost and schedule implications. This is a common challenge when scanning large projects.

9. Are the file formats correct and indexed for performance?

The deliverable must be in the agreed-upon format and structured for efficient use in your design software.

• How to Perform It: Verify that the delivered files have the correct extensions as defined in the SOW, most commonly .RCP for Autodesk workflows or a vendor-neutral .E57 file. Attempt to link the primary project file (e.g., the master .RCP) into a blank Revit project. Note how long it takes to load and how smoothly you can navigate.
• Pass/Fail Criteria: Pass if the file formats are correct and the data loads and performs well in your software. A well-structured project file (like an indexed .RCP) will load much faster than a single, massive unstructured cloud. The data fails if the provider delivered the wrong format or simply gave you a folder of raw, unprocessed scans.
• What to Do If It Fails: Contact the provider immediately and request the data in the specified, structured format. Work cannot begin until the correct deliverables are received. Reviewing different point cloud file formats can help you specify this correctly in future scopes.

10. Are scan positions and panoramic views complete?

The 360° panoramic photos captured at each scanner location are an invaluable tool for the modeling team to resolve ambiguities in the point cloud.

• How to Perform It: In a viewer like Autodesk Recap, enable the visibility of "scan locations" or "mirror balls." This will show an icon at every position where the scanner was set up. Confirm these locations cover the entire project scope. Click on a dozen or so of these icons, especially in mechanically dense rooms, to open the 360° panoramic image.
• Pass/Fail Criteria: Pass if the scan positions are present and the associated panoramic images are clear, well-lit, and correctly oriented. They should serve as a high-resolution virtual visit to the site for your modelers. The data fails if panoramas are missing, corrupt, blurry, or misaligned with the point cloud data.
• What to Do If It Fails: Missing or unusable panoramic imagery is a significant handicap to the modeling team. It increases the risk of misinterpreting complex geometry. The provider should be able to re-export the project with the panoramic data properly included.

Where to go next

Running these ten checks helps ensure your Scan to BIM project starts with a solid foundation. For a deeper understanding of how these issues impact downstream modeling, see the companion guide on point cloud QA for Scan to BIM. To learn what to look for before you even hire a provider, review the guide on how to vet point cloud accuracy before you hire.