Geometry & modeling

Point cloud

Definition

A point cloud is a set of measured coordinates that sample the surface or surroundings of a physical object in three dimensions. The points may also carry colour or intensity, but they do not by themselves define connected faces, exact curves, material volume, or CAD feature history.

Also known as: 3D point cloud, scan point cloud

Updated

What the points store#

Autodesk describes a point cloud as a large collection of points placed in a three-dimensional coordinate system. A scanner or photogrammetry workflow estimates each point's position. Some formats also store colour, return intensity, classification, or scanner information.

The density of points is not the same as accuracy. Capture uncertainty, surface reflectivity, viewing angle, occlusion, registration, and later filtering can all affect the samples. A dense cloud can still contain systematic error or miss a hidden surface completely.

Point cloud, mesh, and CAD body#

A point cloud contains samples without explicit connections between them. A mesh connects vertices into polygons, normally triangles, to form a surface. A B-Rep CAD body represents faces, edges, curves, and topology as engineering geometry.

Converting between those representations requires interpretation. Meshing decides which points belong to one surface and how gaps are closed. Scan-to-CAD work then fits or rebuilds planes, cylinders, curves, and freeform surfaces. The point cloud does not contain the original sketches, dimensions, tolerances, or feature order.

Use in reverse engineering#

SOLIDWORKS Design 2026 accepts point-cloud or mesh files as scan data, then describes preparation, sectioning, surface creation, deviation checking, trimming, knitting, and solid creation.

Preserve the original cloud as measurement evidence. Perform filtering, alignment, meshing, and feature fitting on derived data, and record the coordinate system and assumptions used. The scan-to-CAD workflow explains why an editable replacement model should recover functional relationships rather than merely connect every measured point.

Sources

All glossary terms