A point cloud is a set of 3D coordinates that describe a real place. Each point has a location in space. Many also carry intensity, color, classification, or GPS time. Together they form a measurable record of a site — a building façade, a stockpile, a corridor, a plant, a quarry.
It is not a photograph and it is not a solid CAD model. It is a dense sample of surfaces as they existed at the moment of capture.
How point clouds are captured
Two methods dominate professional work.
Laser scanning / LiDAR measures range with a laser. Terrestrial scanners, mobile mapping systems, and aerial LiDAR all produce point clouds. Density, noise, and coverage depend on the instrument, range, and scan pattern.
Photogrammetry reconstructs 3D points from overlapping photographs. Drone mapping and close-range photogrammetry both produce clouds, often colored from the source imagery. The geometry is derived, not directly measured by a laser.
Both can be accurate. Both can be messy. The quality of the cloud is a function of capture geometry, processing, and how well the coordinate system was controlled.
What the files actually contain
The common interchange formats are LAS and its compressed counterpart LAZ. E57 is widely used for terrestrial scanner packages. ASCII XYZ still appears in older workflows.
A typical LAS/LAZ file stores:
- X, Y, Z coordinates (often in a projected CRS)
- Intensity
- Return number and classification
- RGB when the cloud is colored
- GPS time and scanner-channel fields on more complete datasets
That is a lot of information, and it is stored for software that understands the format. It is not a JPEG. Double-clicking the file on a typical office computer does nothing useful.
Why point clouds are hard to hand off
A finished cloud is often several gigabytes. It may be tiled. It may only make sense in the project CRS. Clients who asked for “the 3D” are rarely in a position to install CloudCompare, Recap, or a GIS stack.
So teams export a screenshot, a flythrough, or a mesh they hope will open. Those products can help. They are not the same as giving someone the spatial record.
A better delivery keeps the cloud viewable in a browser, in the same project as the orthomosaic, model, and report. The specialist file can still exist for the people who need it. The client-facing experience should not depend on them installing a viewer.
When a point cloud is the right deliverable
Use a cloud when the job needs measurable 3D structure that a 2D ortho cannot show: elevations, clearances, stockpile faces, as-built interiors, vegetation, or irregular industrial geometry.
Use a mesh or Gaussian splat when the priority is visual readability of surfaces. Use an orthomosaic when the priority is a true-to-scale map. Serious projects often need more than one of these, attached to the same site.