What is a point cloud?

The short definition

A point cloud is a set of points in 3D space. Each point has an X, a Y, and a Z coordinate, so it sits at a specific spot in the world. Collect millions of them off a real surface and the cloud starts to look like the thing you measured. A building, a road, a quarry wall, a bridge. Nothing is connected. There are no faces or lines like in a CAD model. It is just a very large pile of measured dots, and that is enough to describe a shape in remarkable detail.

Most points carry more than position. Depending on how the data was captured, a single point can also store a color, a measure of how strongly it reflected the sensor, and a label saying what kind of thing it is. More on each of those below.

How point clouds are captured

There are two main families of capture, and they often end up in the same kind of file even though the sensors are different.

• Lidar. A laser scanner fires pulses and measures how long each one takes to bounce back. Time of flight gives distance, and distance plus the sensor angle gives a coordinate. A scanner on a tripod is terrestrial lidar and is used for buildings, interiors, tunnels, and detailed site work. A scanner on a plane, helicopter, or drone is aerial lidar and is used for corridors, large sites, and terrain. Aerial systems pair the scanner with GPS and an inertial unit so every point is tied to a real coordinate as the aircraft moves.
• Photogrammetry. Instead of a laser, you take a lot of overlapping photographs, usually from a drone, and software finds the same features across many images to work out where each point must be in 3D. The result is a point cloud too. It is cheaper to fly and gives natural color straight from the photos, though it is generally less precise than lidar and struggles with featureless or shiny surfaces.

Both approaches produce the same basic thing, a dense set of XYZ points, which is why they share file formats and viewers.

What is in a point: classification, intensity, and RGB

The position is the core of every point. The extra attributes are what make a cloud useful to work with rather than just a shape.

• RGB. A red, green, and blue color for the point. Photogrammetry clouds get this from the photos. Lidar clouds get it when the scanner has a camera or when color is added later from imagery. Color makes a cloud read like a photo you can fly through.
• Intensity. How strongly the surface reflected the laser pulse. It comes from lidar, not photos. Intensity often makes paint stripes, signs, and material changes pop out even when there is no color, which is why surveyors lean on it for things like lane markings.
• Classification. A label on each point saying what it represents, such as ground, building, vegetation, or water. Classification is usually assigned by software after capture. Once points are classified you can do things like show only the bare earth to build a terrain surface, or hide the trees to see the structure underneath.

LAS and LAZ files

Point clouds are almost always shipped as LAS or LAZ. LAS is the standard binary format for lidar data, maintained by the American Society for Photogrammetry and Remote Sensing. It stores all the points along with their attributes and a header describing the file, the coordinate system, and the point format. It is the lingua franca of the industry, and just about every piece of survey and mapping software reads and writes it.

LAS vs LAZ

LAZ is simply LAS that has been compressed. It is the same data, losslessly squeezed down with an open compressor called LASzip, and it usually lands around a fifth to a tenth of the original size with nothing thrown away. A file that is fifty gigabytes as LAS might be six or seven as LAZ. Because the savings are large and the data is identical, most people store and send LAZ by default and only expand to LAS when a tool insists on it. If you see a .las and a .laz side by side, treat them as the same cloud in two wrappers.

How big are these files

Point clouds are heavy because they are dense. Rough numbers, since it depends entirely on point spacing and area:

• A small drone photogrammetry job might be a few hundred megabytes to a couple of gigabytes.
• A terrestrial scan of a building or an intersection often runs ten to fifty gigabytes.
• A large aerial lidar survey can reach into the hundreds of gigabytes or more.

That size is the reason these files are awkward to email, slow to download, and the reason most viewers stream them rather than loading everything at once.

What point clouds are used for

In surveying and the broader AEC world, the cloud is rarely the final product. It is the raw measurement that everything else is built from.

• Surveying. Topographic surveys, as-builts, and existing conditions. Crews extract breaklines, surfaces, and features from the cloud instead of walking every point with a rod.
• Construction. Comparing what was built against the design model to catch errors, tracking earthwork volumes between flights, and documenting a site at a moment in time.
• Architecture and engineering. Starting a renovation or retrofit from an accurate record of the building that exists, rather than from drawings that may be decades old and wrong.
• Inspection and documentation. Bridges, towers, interiors, and anything where you want a measurable 3D record you can come back to later.

How you view and share them

Because a point cloud is 3D and large, you need software to look at one. The traditional tools are desktop applications: Autodesk ReCap and Civil 3D, Bentley products, and the free and capable CloudCompare. They are powerful, they run locally, and they expect you to install and learn them. That is fine for the people processing the data.

The problem shows up when you need to share the cloud with someone who does not have any of that software, which is most clients. The modern answer is a browser-based viewer that streams the cloud over the web. The file stays on a server, the viewer pulls only the points currently on screen, and the cloud opens in a normal browser with no install. That makes a fifty gigabyte scan something you can hand to a client as a link.

That browser approach is what ViewerFX does for LAS and LAZ files. If you want to see what a streaming point cloud actually feels like, the free viewer opens one in your browser with no signup. Drop in a file, orbit it, color it by elevation or classification, and pull a measurement off it. It is the fastest way to understand what all those millions of points add up to.