Georeferenced 3D Capture on iPhone: Meshes, Point Clouds, and Photo Scans

A phone can produce several very different objects that all get called a “3D scan.” A LiDAR mesh, a raw point cloud, and a photogrammetric reconstruction have different strengths, failure modes, and processing costs. For field documentation, the capture method is only half the problem. The model also needs a defensible answer to: where is it?

Choose the capture for the question

None is universally “most accurate.” The right choice depends on scale, surface texture, lighting, device, how quickly the result is needed, and whether the deliverable needs geometry, appearance, or both.

Anchor before capture, not after

A local 3D model begins in an arbitrary coordinate frame. If the operator later drags it onto a map by eye, the result is placed, not surveyed. OpenSurvey3D starts a 3D capture from a project observation. The app locks the capture origin to that asset and stores the position source and expected accuracy before the camera begins.

The preflight states the coordinate system, source, position accuracy, pole or camera offset, and calibration state in one place. If the capture cannot be meaningfully georeferenced, that is a blocking condition. If conditions are merely weak — poor accuracy, thermal pressure, low storage, stale corrections — the warning remains visible in the capture record.

Position is not the entire transform

A 3D asset needs translation, rotation, scale, and a vertical policy. GNSS establishes a global origin, but the device still needs a heading and a relationship between the antenna and camera. Useful records include:

• the anchor observation and its horizontal and vertical accuracy;
• the local-to-project transform and the coordinate frame it targets;
• heading source and heading accuracy;
• camera-to-antenna offset for a handheld or pole-mounted rig;
• fit RMS when multiple observations refine the alignment;
• whether height is measured, terrain-snapped, or unknown.

This metadata prevents a polished model from implying more positional certainty than its anchor can support. A scan tied to a phone-GPS observation remains meter-class even if its local surface is visually crisp.

Coverage is a field-quality question

Capture problems are cheapest to fix before leaving. A mesh can have holes behind an obstruction. A point cloud can thin out on dark or reflective surfaces. A photo scan can fail because one side of the object has too little overlap. OpenSurvey3D records live health and ends the capture with a coverage review so the operator can inspect the result while the subject is still present.

Photogrammetry jobs are resumable because source images are the irreplaceable part of the record. Point-cloud captures can optionally retain posed frames — color, depth, camera pose, and the GNSS fix behind the frame — so a different processing engine can reconstruct the capture later instead of being limited to today's output.

The project model connects the evidence

An isolated scan viewer answers “what did we capture?” A site model answers “how does it relate to everything else?” OpenSurvey3D builds a RealityKit scene for a chosen working area from terrain heights and draped imagery, then adds project points, linework, routes, labels, elevation profiles, and anchored scans.

The working area is explicit because resolution and download size are linked. A small site can use sharper terrain and imagery than an entire county. Once built, the model stays cached for offline use, can grow as the project grows, and can export terrain geometry as OBJ.

What an anchored phone capture is good for

• Documenting existing conditions at an asset on a specific date.
• Providing visual context around measured points, utilities, or damage.
• Comparing an as-built surface with design or earlier capture geometry.
• Carrying a route, profile, scan, and survey observations in one project scene.
• Preserving source evidence for higher-quality offline processing.

It is not automatically a boundary survey, a certified deformation analysis, or a replacement for controlled terrestrial scanning. The usefulness comes from making the capture's frame and limitations explicit.

Return to the capture through AR

Georeferencing also makes the record useful after capture day. OpenSurvey3D can show nearby project assets as pins in the live camera view, guide the operator toward one target by direction and distance, then reopen the photos or 3D scan attached to its record. The AR pin is a finding aid, not a precision stakeout result: it inherits current GNSS, stored-coordinate, heading, and tracking uncertainty.

See Finding Surveyed Assets in AR for the coordinate-to-camera workflow and its practical limits.

A practical field sequence

1. Set the project coordinate system and confirm the positioning source.
2. Measure or select the asset that will anchor the capture.
3. Enter the camera/antenna offset and review preflight warnings.
4. Capture with the mode that matches the deliverable.
5. Review coverage, holes, alignment, and source-image completeness before leaving.
6. Open the scan in the project model and record any independent check shots.
7. Export the project bundle with the QA record and source material required for handoff.