Point cloud

A point cloud is a large set of 3D coordinates (x, y, z) that represent the surface of a scanned object. Each point corresponds to one location on the object's surface. Industrial scanners capture millions of points per minute, so the data for a single part can contain tens of millions of points.

The point cloud is the primary output of 3D scanning. In its raw form it looks like a "cloud" of points in the shape of the scanned object. To be useful, the data has to be processed: noise is filtered out, multiple scans are aligned into a common coordinate system (registration), and a polygon mesh is generated or CAD geometry is reconstructed directly.

Software for working with point clouds includes Geomagic, PolyWorks, CloudCompare (open source) and modules built into CAD systems (for example ScanTo3D in SolidWorks). These tools let you visualize and edit the data and convert it into a polygon mesh or into a surface or solid CAD model.

In quality inspection, the point cloud is compared with the nominal CAD model. A color deviation map immediately shows where the actual part deviates from the design. This comparison is used both in first article inspection (FAI) and in inspection during series production.

You work with a point cloud in every 3D scanning project. It is used directly mainly for comparing an actual part with its CAD model, for measuring dimensions and distances on the scanned object, and for documenting the as-is condition (for example before a machine is repaired or modified).

For large objects such as production lines, factory halls or bulky equipment, a point cloud from a laser scanner serves as the basis for fitting new equipment into the existing space, without having to model the entire object in CAD.

A point cloud contains noise, reflections, gaps (shadowed areas, holes) and, when scans are combined, registration errors. Before measuring or comparing, the data must therefore be cleaned, and you should check how accurately the individual scans are aligned.

The data is often very large. Agree with the provider on the format (E57, PLY, XYZ, or directly an STL mesh) and on a point density that suits the purpose. A sparser cloud is enough for fitting in new equipment, while inspecting fine shape details needs a denser one. Do not confuse a point cloud with a CAD model: you cannot create a drawing or a CNC program directly from it.