Modeling from Point Clouds | Why We Never Import Directly into CAD

No one asks how we model—until it’s too late
Requests for proposals usually focus on the perimeter, the number of stories, floor plans, elevations, and whether a 2D or 3D model is needed. Some also ask about the desired level of detail.

But one key question almost always goes unnoticed: How is the modeling actually done?
It is precisely this question that determines everything later on—especially when dimensions are incorrect, planning comes to a standstill, or the model is difficult to work with further.

The Obvious, but Risky Approach
Many people today follow a seemingly logical principle: The point cloud is cleaned up, then loaded directly into CAD, and break edges are visually estimated.

What sounds reasonable is often riskily inaccurate in practice. Additionally, there’s the question of whether your own computer can even handle the full data set efficiently.

This approach carries several risks:
When cleaning up the point cloud, the very information needed to accurately identify the actual break edges is often removed—especially in the case of deformed or complex older buildings. Those who work directly in CAD rarely hit the exact break edge, but rather a random point somewhere nearby. This is more a matter of estimation than precise work.

If the point cloud is internally displaced, these errors are carried over directly into the model unnoticed—with potentially serious consequences for the design.

We do it differently
Before we even open a CAD file, we perform a manual displacement analysis of the point cloud data. The reason is simple: currently available point cloud registration software is unable to reliably detect such internal errors. It requires a critical eye—experience rather than automation.

Furthermore, we do not “clean up” our point clouds. What is often referred to as “cleaning up” in many workflows actually means deleting reality. For us, it is clear that every piece of information—no matter how disorganized it may seem at first glance—holds potential value for modeling.

Point by Point Instead of Estimates
That’s why we never import the point cloud directly into CAD. We’ve developed our own tool that allows us to click, point by point, on the actual break edges—not on noise, not on estimated lines, but on real geometries. Only these relevant, precisely selected points are automatically transferred to the CAD system—in the software and format our customers need. Not the entire point cloud. Just what matters. Point by point.

Frequently Asked Questions

Why isn’t a cleaned-up point cloud plus a visual estimate in CAD sufficient?
During the cleaning process, the very information needed for the precise detection of break edges is often lost. A visual estimate inCADrarely pinpoints the actual edge with precision.

What is a manual displacement analysis?
An additional, software-independent check of the point cloud for internal displacements between scan positions—before any modeling takes place.

Why doesn’t VDE clean up point clouds?
Because even seemingly disordered data can contain valuable information for modeling that would be lost during cleanup.

How does VDE ensure that break edges are captured correctly?
Using a proprietary tool that maps each point individually to the actual geometry, rather than transferring the entire point cloud into CAD.