The physical world is increasingly being captured in 3D. Infrastructure, natural environments, underground spaces, heritage sites — all being documented with a precision that would have seemed extraordinary a decade ago.
Here’s a grounded explanation of what 3D mapping is, how the main technologies work, and what the outputs actually look like.
What 3D mapping means in practice
3D mapping is the process of capturing and representing physical spaces as three-dimensional digital datasets. The output can take many forms:
- A point cloud: Millions of georeferenced points in XYZ space, each potentially carrying additional attributes (colour, intensity, classification)
- A mesh model: A surface representation made of polygons, usually textured with photographic imagery
- A Digital Surface Model (DSM) or Digital Terrain Model (DTM): A raster grid representing elevation across an area
- A 3D orthomosaic: A top-down image stitched from aerial photography with accurate geospatial reference
All of these are representations of the same physical reality — just at different levels of abstraction and suited to different downstream uses.
The main capture technologies
LiDAR (Light Detection and Ranging)
LiDAR is an active sensor technology. It fires pulses of laser light at the environment and measures the time it takes for the light to return. From this, it calculates the precise distance to each reflection point.
Modern LiDAR sensors fire hundreds of thousands of pulses per second, building up dense point clouds in real time. They can be:
- Terrestrial: Tripod-mounted scanners for building interiors, infrastructure, underground spaces
- Mobile: Vehicle-mounted or backpack systems for roads, tunnels, urban corridors
- Airborne: Fixed-wing or helicopter-mounted for large-area terrain mapping
- UAV-mounted: Drone-mounted LiDAR for medium-scale surveys
What LiDAR produces: Dense point clouds in LAS or LAZ format, with XYZ coordinates, intensity values, and often return counts (useful for vegetation penetration analysis).
Photogrammetry
Photogrammetry derives 3D geometry from overlapping 2D photographs. Computer vision algorithms identify matching features across multiple images taken from different positions and solve for both camera positions and the 3D structure of the scene simultaneously — a process called Structure from Motion (SfM).
For aerial work, this typically means:
- A drone captures hundreds or thousands of overlapping images
- Software (RealityCapture, Metashape, Pix4D) processes them into a dense point cloud and mesh
- GCPs (Ground Control Points) measured with GPS/GNSS tie the model to real-world coordinates
What photogrammetry produces: Dense point clouds, mesh models, textured 3D models, orthomosaics, DSMs.
Key difference from LiDAR: Photogrammetry is passive (works with reflected light, requires texture and contrast) and generally cheaper. LiDAR is active (works in darkness, penetrates vegetation) and generally more accurate for bare-earth capture.
360° capture
360° cameras capture immersive panoramic imagery of spaces. Unlike LiDAR and photogrammetry, 360° capture doesn’t produce geometric 3D data — it produces visual 3D experiences (spherical images and video that you can look around in).
For documentation purposes, 360° capture complements geometric data: while a point cloud captures the geometry of a space, 360° imagery captures the visual experience of being inside it.
SLAM (Simultaneous Localisation and Mapping)
SLAM is a technique used by mobile mapping systems (backpacks, handheld scanners, autonomous robots) to build a map while simultaneously tracking their own position within it. It’s what allows a backpack LiDAR system to capture a building interior without GPS.
SLAM systems like Leica BLK360, Navvis M6, and GeoSLAM ZEB produce point clouds suitable for as-built documentation, heritage recording, and underground mapping.
The data problem
3D mapping technology has advanced dramatically. LiDAR sensors that cost $75,000 five years ago now cost $5,000. Photogrammetry software that required a compute cluster now runs on a high-spec laptop. Drone flights that took a day’s planning can be done in an hour.
The technology to capture has democratised. But the data problem remains:
3D datasets are enormous. A single corridor LiDAR scan can be 10-50 GB. A drone survey of a medium-sized construction site might produce 5-20 GB of processed outputs.
The tools to view them are specialist. CloudCompare, QGIS, RealityCapture, ArcGIS Pro — none of these are applications a non-specialist engineer or project manager wants to install.
Delivery hasn’t kept up. The industry standard for delivering spatial data to clients is still: compress the files, upload to Dropbox, email the link. This is not a professional delivery mechanism for data that cost thousands of dollars to capture and is critical to decision-making.
The delivery solution
Modern 3D data delivery uses streaming formats (3D Tiles for meshes, Potree format for point clouds) that break large datasets into chunks, loading only what’s visible in the browser viewport.
The result: a 20 GB point cloud can be explored in any browser, on any device, without downloading the full file. The viewer looks and feels like Google Earth — because it uses the same underlying technology (WebGL, CesiumJS, Potree).
For surveyors and drone operators, the practical implication is: your clients can receive a link, open it in their browser, and explore their survey data at full fidelity — without any software installation.
That’s what professional spatial data delivery looks like in 2025. A survey captured last week should be accessible in a browser this week. That’s the gap Swyvl is built to close.
3D mapping is rapidly becoming a standard part of how infrastructure is documented, monitored, and managed. The technology to capture is solved. The challenge — and the opportunity — is in the delivery layer.