Phone-First Photogrammetry: Build a 3D Site Model with Your Smartphone

Written By Kristina Delvallee

Photogrammetry turns overlapping photos into a measurable 3D model. Under the hood, Structure-from-Motion estimates camera positions from feature matches; Multi-View Stereo densifies those points into a surface you can read for tool marks, reuse, or later repairs—the kind of evidence your eye catches on site but is hard to share.

What you’ll need (no fancy gear)

  • Your phone. Any recent iPhone/Android works. Lock focus/exposure and avoid HDR to keep frames consistent. Consistency > megapixels. (General capture tips emphasize manual/consistent settings.)

  • Free desktop software (pick one):

    • Meshroom (AliceVision) — GUI, fully free.

    • COLMAP — GUI/CLI, research-grade; superb results.

    • OpenDroneMap / WebODM — great for outdoor/large scenes.

  • Optional: a paper ruler or scale card for later true-to-scale models.

Capture workflow (phone only)

Goal: high overlap, stable exposure, all angles.

  1. Prep the subject

    • Favor matte, textured surfaces; avoid glass/metal if you can. (Reflective/low-texture surfaces are challenging.)

  2. Light & settings

    • Use soft, even light; kill harsh speculars. Lock AE/AF. (Beginner guides stress consistent lighting and manual settings.)

  3. Shoot path

    • Walk a slow orbit at chest height, then a slightly higher ring, then top-down passes.

    • Aim for 60–80% overlap frame-to-frame; take more than you think you need.

  4. Quality control

    • Blur kills models. Brace elbows, keep shutter fast. If indoors (e.g., museums), remember many spaces ban tripods/flash.

Process (free tools)

Option A — Meshroom (easiest GUI)

  1. Drop images → Start.

  2. Inspect the graph as it runs (FeatureExtraction → SfM → DepthMap → Meshing → Texturing).

  3. Export the textured mesh (OBJ/GLB). (Meshroom tutorial + capture guidance here.)

Option B — COLMAP (more control)

  1. Feature Extraction → Matching → Reconstruction (sparse).

  2. Dense (MVS) → point cloud → mesh via Poisson/screened Poisson.

  3. Export and, if needed, unwrap/decimate in Blender for the web. (Official tutorial explains SfM→MVS→surface steps.)

Option C — OpenDroneMap / WebODM (outdoors)

Batch big scenes (façades, courtyards) and produce orthos/meshes; ODM docs include overlap presets.

Make it to scale (so measurements are meaningful)

Add a known-distance to the scene (printed scale bar, ruler) and set scale during/after processing. Cultural Heritage Imaging’s scale-bar guidance is the standard quick fix; Metashape docs show the same principle if you need a UI example.

Read your model (the evidence lens)

  • In-situ vs. moved. Base contact, bedding planes, and mortar seams tell you if a block or statue is original to context.

  • Reuse & repair. Chisel picks, patch textures, and toolpaths pop in a shaded model.

  • Water & wear. Rills and polish show flows and footfall—key to your “lostness” patterns.

Common pitfalls (and quick fixes)

  • Too little overlap → holes. Shoot slower; keep 60–80% and vary height.

  • Auto-exposure flicker → mismatched colors. Lock exposure/focus.

  • Shiny/featureless surfaces → tracking fails. Add matte tape/markers nearby or choose a different subject.

  • Heavy models → slow embeds. Decimate and re-bake texture; Sketchfab’s optimization tips help.

Ethics & permissions (non-negotiable)

  • Museums/sites. Policies vary: The Met allows non-commercial handheld photography but restricts use; always check first. The British Museum explicitly allows handheld cameras and even “3D imaging software” unless posted otherwise. No tripods without permission; be respectful.

Quickstart checklist

  • Subject chosen (matte/texture)

  • AE/AF locked, HDR off

  • Two orbits + top-down passes

  • ≥60–80% overlap

  • Include scale (ruler/scale bar)

  • Process: Meshroom or COLMAP

  • Export GLB (web) + OBJ (archive)

  • Note permissions/credits for where you captured

Further reading & tools

Kristina Delvallee
Previous

UNESCO 101 for Travelers: What It Is (and What It Isn’t)
Next

Toponymy as Evidence: How Place-Names Preserve Lost Landscapes