Construction teams have relied on cameras, notepads, and separate survey crews to document job sites for decades — a slow, fragmented process that doesn't scale. Torch Solutions built a WebGIS platform that turns any LiDAR-equipped iPhone into a full site-documentation and measurement tool, producing GIS-ready 3D models without the need for specialized survey equipment. Here's how we approached the engineering behind it.
The idea behind our WebGIS platform started with a simple observation: field engineers were still documenting construction sites with a camera, a notepad, and sometimes a separate survey crew for anything requiring real measurements. Modern iPhones have had LiDAR sensors since the Pro models launched — the hardware to fix this was already sitting in people's pockets. The software to make it useful wasn't.
The core technical challenge going in was deceptively simple to describe: capture a LiDAR point cloud and geotagged photographs at the same time, in one scanning session, without either stream degrading the other. In practice this meant working closely with ARKit and RealityKit to keep spatial tracking stable while the camera was also recording overlapping site imagery, and doing it all through Metal so the live point-cloud preview didn't lag on-device.
Get this wrong and you end up with a beautiful 3D model that's misaligned with the photos taken during the same walk, which is worse than useless for anyone trying to measure a trench or verify a foundation pour. We spent a disproportionate amount of time on this synchronization problem relative to how invisible it is in the final product — which is usually a sign you got it right.
Construction sites are not known for great connectivity. A scan session can generate a large package of point-cloud and image data, and uploading that over a spotty rural or basement-level connection isn't something you can just wrap in a retry loop and hope for the best.
The approach that worked was keeping captured data fully on-device until a stable connection was available, then syncing with proper retry behavior rather than forcing an upload-or-fail moment in the field. It's a less exciting engineering decision than the LiDAR pipeline, but it's the one that determines whether field engineers actually trust the tool enough to rely on it instead of falling back to old habits.
Capturing the data is only half the platform. On the backend, a Django REST Framework service picks up each scan and runs it through an AliceVision photogrammetry pipeline, producing 3D models in USDZ and GLB formats alongside georeferenced GeoTIFF outputs that plug directly into tools like QGIS and ArcGIS that site teams already use.
This mattered more than it might sound. A lot of "innovative" field-capture tools produce output that only lives inside their own proprietary viewer, which means someone on the GIS or surveying side still has to redo the work in a format they can actually use. Making the raw outputs interoperable with existing GIS workflows was as much a product decision as a technical one.
A single scanning session can now cover site features — trenches, foundations — across distances up to around 100 meters, with distance, area, and volume measurements available directly from the web dashboard afterward. Field engineers get a workflow that goes from phone in hand to measurable, GIS-ready output without a specialized survey crew for every routine capture.
The broader takeaway from this build: spatial computing projects live or die on the boring parts — sync reliability, offline behavior, and format compatibility with whatever tools your users already trust. The LiDAR capture is the part people notice first. The plumbing around it is the part that determines whether they keep using it.
iPhone LiDAR scanning is accurate enough for routine site documentation, progress tracking, and preliminary volume or area estimates — typically within a few centimeters under normal conditions. For legal surveys or final as-built certification, teams still pair it with traditional survey methods, but for day-to-day trench and foundation checks it removes the need to schedule a separate crew.
Scans are processed into USDZ and GLB 3D models plus georeferenced GeoTIFF outputs, so they open directly in QGIS, ArcGIS, and other GIS tools your team already uses — no proprietary viewer required.
Yes. Scan data stays on the device until a stable connection is available, then syncs automatically with retry handling. This is designed specifically for construction sites with unreliable rural or below-grade signal.
A single scanning session can capture site features like trenches and foundations across distances of up to around 100 meters, with distance, area, and volume measurements generated automatically once the scan syncs to the web dashboard.
Want to see the full build in detail, including the architecture decisions and results? Read the complete WebGIS platform case study.
