CloudCompare

What CloudCompare does

CloudCompare is a desktop application built for very large three-dimensional point sets: laser scans from terrestrial scanners, drone photogrammetry output, LiDAR strips, and the polygon meshes that usually accompany them. It opens files with tens of millions of points without falling over, keeps each scan as a separate entity in a tree, and gives every entity its own visibility toggle, scalar field colouring and transformation history. Typical work includes comparing an as-built scan with a reference CAD model, measuring deformation between two epochs of the same site, cropping out a region of interest, and exporting a cleaned mesh for downstream work.

How the workflow usually runs

Most projects start by importing the raw files and immediately cutting the data down, because a full scan of a building is far more than any analysis needs. The segmentation tools use an interactive polyline or a simple box to isolate a facade, a floor or a single object, and everything removed stays available in the database rather than being deleted outright. Registration comes next: pairs of scans are aligned either by picking matching point pairs by hand, by choosing equivalent triples of points, or automatically with the ICP implementation, which refines an approximate alignment until the overlap converges. A fine registration step reports the final residual error so you can judge whether the alignment is good enough for measurement.

Once clouds are aligned, the analysis tools do the interesting part. Cloud-to-cloud distance computes a signed or unsigned deviation field, which you can display as a colour ramp and threshold to highlight anything outside tolerance. Cloud-to-mesh distance does the same against a triangulated reference. Cross-section extraction cuts a slice through the data and can unroll it into a flat profile, which is how flatness, ovality and floor level are checked on real projects. Volume computation, roughness statistics, density estimation and normal computation round out the toolbox, and every result can be exported as a scalar field, a mesh or plain ASCII coordinates.

Settings, performance and limits

Performance is dominated by memory rather than CPU. The application keeps points in memory with an octree for spatial queries, so an eight-gigabyte machine handles single scans comfortably but struggles with a dozen unmerged aerial tiles at once. The subsampling and octree-level tools exist precisely for that reason, and lowering the displayed point size or hiding entities gives an immediate speed-up on weak graphics hardware. Commands can be recorded and replayed, and the command-line mode allows batch conversion, which matters when a project has hundreds of files. The interface is dense and clearly designed by engineers: menus are long, dialogs expose many parameters, and there is no guided wizard to hold your hand. Some advanced plugins, notably the photogrammetry-oriented ones, add extra dependencies and are shipped separately.

Who should choose something else

If your goal is architectural modelling rather than inspection, a BIM or CAD package will be far more productive than a point-cloud viewer. If you need to produce a solid parametric part from scan data, reverse-engineering software with surface fitting will save you days. And if you only ever look at small meshes from a 3D printer, a slicer preview or a lightweight mesh viewer is enough. CloudCompare earns its place in survey, heritage, tunnelling, plant and quality-control work, where the questions are about deviation, alignment and volume rather than about design intent.