What Is a Point Cloud?

If you're taking a 3D laser scan of your site’s as-is condition and turning that information into a 3D CAD model to estimate, detail, or make renovations to MEP projects, you might wonder what a point cloud is. 

To understand a point cloud, it’s important to learn how it’s created in the first place.

It actually takes three steps to get a site from scan to model.

1. Scan the area with a 3D laser scanner

2. Import the point cloud into modeling software to visualize the area

3. Export the point cloud and import it into a CAD/BIM system

First, use a 3D laser scanner to scan your area’s measurement. Next, import the point cloud the scanner creates (and that you don’t initially see) into point-cloud modeling software. The software lets you visualize and model the point cloud, which at that stage looks rather like a pixelated, digital version of your site. In the last step, you’ll export your point cloud from the software and import it into a CAD/BIM system.

In one Constructible blog, Aric Stott gave a primer on how to create a 3D CAD model from a 3D laser-scanner, using the example of an Olympic swimming pool and its surroundings.

Many new to laser scanning ask: what’s a point cloud and why is that middle step (importing) needed? The short answer is, scanned information can’t be directly imported to CAD software. The long answer calls for a definition of the point cloud.

What is a point cloud?

When you take a scan, the laser scanner records a huge number of data points returned from the surfaces in the area you’re scanning. These can include walls, windows, ductwork, steel structures, etc. In the case of the swimming pool, Stott’s team scanned the data points and even included the flags hanging above the pool and the plastic dividers used to mark off swimming lanes.

Scan from Aric Stott’s project, showing a swimming pool and flag details

The data points usually exist along the x, y, and z coordinates within the 3D scanned space.

From Sky to Art to Software: Point Clouds, De(mist)ified

A cloud is a 3D mass made up of small droplets, crystals, water, or various chemicals. In the same way, a point cloud is a huge number of tiny data points that exist in three dimensions. If you could spit those points out of a scanner they’d appear as a cloud you could walk within.

Take the video below as an example. This is what it feels like to fly through a point cloud.

Point Cloud video created by David Hyland

So the point cloud that the laser scanner captures is an accurate as-built of an object or space. It’s saved in the form of a very large number of points that cover surfaces of an object. And a point cloud defines a space by recording the points that cover the surfaces within that space.

For a detailed look at scanning best practices, check out this Scan-to-BIM ebook.

Or, think of a point cloud as a painting made through pointillism, a technique in which the artist applies a pattern of small, distinct dots of color—often no bigger than the tip of a paintbrush—applied to a canvas, eventually creating an image. You’ve probably seen a popular example of that type of work: “A Sunday Afternoon on the Island of La Grand Jatte”, created by Georges Pierre-Seurat in 1884. The painting depicts men and women in clothing of the day—including hats, bustles, and parasols—relaxing on a verdant bank of the River Seine as sailboats drift across the water.

A Sunday Afternoon on the Island of La Grand Jatte, Georges Pierre-Seurat, 1884

If you were to take a closer look at the painting, you would see each individual “point” that makes up “pointillism.” The same goes for a point cloud.

Closer look at A Sunday Afternoon on the Island of La Grand Jatte, Georges Pierre-Seurat, 1884

The “point” here is many tiny dots (points) create a whole, as seen in Stott’s point cloud scanned from the swimming pool. You can see the image is made of individual points, but you can pick out the objects surrounding the pool and get the scale of the room itself.

To get the pool point cloud, Stott and his team scanned the room from six different locations and then imported the scanned files into Trimble RealWorks point-cloud modeling software. The software easily returns a visual image of the of the point cloud.

Many Clouds Make a Whole

Because you can’t obtain points on the surfaces, the scanner can’t “see” you need to take scans from multiple locations to get the complete view of your target area. With the Trimble RealWorks software, Stott’s team simply and automatically “stitched” the six-point clouds returned from six scans taken from various locations to create the image of the point-cloud pool.

Once imported into the software, the point cloud can be parsed, manipulated, and modified to suit a user’s needs. It also provides point cloud data management, analysis, and advanced modeling.

To work with the information as a CAD model, you’ll export the point cloud from the software and import it into a CAD modeling program where the data is turned into 3D solids and surfaces. This part of the process can also be done, for example, in Trimble Realworks.

In the past, point clouds were too big to import directly into 3D modeling software.  Today's CAD software, however, can handle very large data sets.

While construction companies may realize the value 3D laser scanners have in creating CAD or Revit models they can work with, employees may not realize the vital second step in getting from scan to model: the point cloud. But the next time they’re looking to create an estimate on an existing project for which no blueprints exist, they could very well thank the clouds—and not the clouds overhead.