What Is a Point Cloud? Guide 2026 (LiDAR, Formats, Uses)

Definition: a point cloud, technically

A point cloud is a discrete set of data points in space. Each point position has a set of Cartesian coordinates (X, Y, Z). Points may also carry additional information: RGB colors (when the scanner has a camera), surface normals (the direction the surface faces at that point), intensity, timestamps, classification (ground vs vegetation vs building, for outdoor LiDAR). Source: buildingSMART community and the Wikipedia reference page on Point Cloud.

Two notes for newcomers. First: a point cloud is not the same thing as a 3D model. A model has surfaces, edges, semantics. A point cloud is just the dots. You can convert dots to surfaces ("meshing") or compare dots to a designed model ("scan vs BIM", called "as-built"), but the raw cloud is geometry without semantics. Second: do not confuse the 3D point cloud (LiDAR, photogrammetry, AEC, surveying, industry) with the 2D scatter plot used in statistics, also called "nuage de points" in French. Same name, completely different topic.

How point clouds are created

1. LiDAR (laser scanning)

A LiDAR scanner emits a laser beam, measures how long the beam takes to return, computes the distance, and records a point. Repeat that millions of times per second, sweeping the field of view, and you get a point cloud. Static terrestrial scanners (used in surveying and AEC) deliver millimeter accuracy. Mobile mapping scanners (mounted on a backpack, a vehicle, a drone, an autonomous robot) trade some accuracy for speed and coverage.

Brands you will encounter on professional sites: FARO, Leica (Hexagon), NavVis, Riegl, Trimble, Viametris, Matterport, GeoSLAM. Each brand has its own scanner range, its own native file format and its own workflow, which is why a platform that handles them natively (without manual format wrangling) saves real time on mixed-fleet operations.

2. Photogrammetry

Take many overlapping photos of a scene from different angles, run them through photogrammetry software (Agisoft Metashape, RealityCapture, ContextCapture, Pix4D, OpenDroneMap), and the software triangulates 3D points from the images. The result is a colored point cloud, sometimes denser than LiDAR, sometimes less accurate, depending on lighting and surface texture. Drone photogrammetry is now the dominant method for large outdoor sites (quarries, agriculture, archaeology, infrastructure).

3. Other sources (depth cameras, smartphones, simulation)

Depth cameras (Intel RealSense, Microsoft Kinect generations, Apple's iPhone Pro / iPad Pro LiDAR sensor), structured-light scanners and even simulated point clouds (in Unity, Unreal, robotics training) all produce data of the same family. The geometry conventions are the same; the accuracy and the file size differ massively.

Common point cloud file formats

• E57: open ASTM standard (E2807). The default interchange format across the industry; supported by most scanners and most software. Good first choice when you do not control downstream tooling.
• LAS: open standard from the American Society for Photogrammetry and Remote Sensing (ASPRS). The reference format for LiDAR data, particularly outdoor / aerial.
• LAZ: a compressed LAS (typically 80 to 90 percent smaller). Same content, much lighter to ship. Open and widely supported.
• RCS / RCP: native Autodesk Recap formats. Read and written natively by ATIS.cloud without an Autodesk license required, which is unusual on the market.
• LGSx: proprietary Leica Hexagon format. The Hexagon ecosystem is migrating progressively from "Leica" to the "Hexagon" naming.
• PLY, OBJ, PCD, PTS, XYZ: legacy or research formats, still encountered, less standard in production AEC workflows.

What professionals do with a point cloud

• Surveying and topography: deliver as-found drawings, volumes (stockpiles, excavations), DTM and DSM
• AEC and BIM: scan an existing building, compare scan vs BIM model (called "as-built"), feed scan-to-BIM workflows
• Industry: document a plant, monitor structural deformation, plan retrofits, train operators in VR / digital twin environments
• Heritage and archaeology: archive a monument with millimeter accuracy, plan a restoration, build digital exhibits
• Infrastructure: bridge inspection, highway maintenance, railway clearance checks, tunnel monitoring
• Forestry and agriculture: forest biomass estimation from aerial LiDAR, plant counts, terrain analysis
• Drone and surveying SaaS: serve scans to clients via secure link, with measurement and annotation in the browser

How to view and share a point cloud

Once a scan is captured, two questions decide what tooling fits: how big is the file, and who needs to see it. For desktop work on small to medium files (LAS, LAZ, E57), CloudCompare (open source) and Autodesk Recap (commercial) are common. For large files or when clients need to look at the scan in a browser without installing anything, a managed SaaS handles the streaming, the access control and the format conversion in one place. ATIS.cloud is the platform we build for that exact use case: scans up to 500 GB per file (up to 5 TB on Enterprise), 8 native scanner brands, formats E57, LAS, LAZ, RCS, RCP, LGSx, secure link sharing, sovereign hosting in 22+ countries, 14-day free trial without credit card.