New blog post! Using 3D scan data created with my phone as reference imagery for underground mapping of subway stations in #OpenStreetMap Check it out! wilkinson.graphics/blog/2025-10...

Mapping Underground Structures...
Mapping Underground Structures with 3D Scans

My recent dive into indoor mapping with OpenStreetMap using photogrammetric references.

Scalable Aerial Imagery Generation from Phone Lidar and 360° Point Clouds

Image Source: Jake Coppinger

OpenStreetMap
@tordans @wilkinson.graphics Yeah! I've read that 2022 post by Jake, it's part of what inspired me to do this!

@hank @tordans @wilkinson.graphics That’s wonderful to hear, thanks!

Yes - I haven’t cracked into OpenDroneMap dev yet but I _think_ adding some tooling around https://github.com/OpenDroneMap/odm_orthophoto would be the go for automation.

GitHub - OpenDroneMap/odm_orthophoto: ODM Orthophoto Module - Rasterize Textured 3D Models to PNG

ODM Orthophoto Module - Rasterize Textured 3D Models to PNG - OpenDroneMap/odm_orthophoto

GitHub
@hank @tordans @wilkinson.graphics
A few more breadcrumbs - WebODM generates intermediate depthmap images. In theory, you could patch the code so that you supply bitmap depthmaps from the iPhone lidar sensor for <=5 metre depth captures - I haven't had a crack at that yet.

@hank @tordans @wilkinson.graphics

Another issue with phone photogrammetry is most apps only embed the GPS data. Ideally, you'd have the inferred GPS + IMU fused data - the LiDAR tracking apps certainly show this is possible.
This is exactly what PIX4Dcatch seems to do - does GPS + IMU sensor fusion, puts those calculated coordinates in, then sends that off to their proprietary photogrammetry. If you export the images before upload, you can input into WebODM - and get stunning results!

@hank @tordans @wilkinson.graphics

Problem is - you need a paid licence to capture and export images at USD$1590 per year after a trial!

I've been thinking perhaps by recording all the iPhone sensors using the SensorLog app while taking photos on a 5 second timer for example, and then doing some fancy sensor fusion maths on the SensorLog output - you could get super detailed coordinates into images for better WebODM stitching.

Anyway - keen to follow along!

@jakecoppinger @hank @wilkinson.graphics did you see https://www.youtube.com/watch?v=N92rACDMRak with https://store.kartaview.org/
They have ideal dataset if points and colors to this but (same as Mapillary) are focussed on the AI part of it and not the (craft) mapping.
Beyond Street Imagery: Crowdsourcing LiDAR ~ Alex Ilisei & Philipp Kandal

YouTube
@tordans @jakecoppinger @wilkinson.graphics Woah! This I hadn't seen. Making your own hardware for this is really impressive.

@jakecoppinger In my experience ODM (or at least WebODM) really doesn't like ground level imagery.

I nearly always wish it had a Hugin-like interface to manually assign some matching points for an initial alignment before letting it try fill in the gaps and refine.

@hank @tordans @wilkinson.graphics

@wilkinson.graphics good thing I don't have access to a device with LiDAR. That would be just one more source of data I don't have time to pour into my #OpenStreetMap addiction. 😉

Thanks for sharing.

@wilkinson.graphics great post! You could try https://osminedit.pavie.info/, which helps with indoor tagging.
OsmInEdit

Simple indoor editor for OpenStreetMap. Let's map inside buildings, the easy way !

@h______ Should have included this in a footnote, but I’ve used this one before. It’s pretty good, but it enforces strict rules about your indoor structures being within the footprint of the building you want to edit, and I didn’t want to figure out what that was exactly as part of this scope of work. I also didn’t want to figure out if that is well supported for underground structures like this one that extend far beyond the building you see above ground.
@wilkinson.graphics this is something I've been wanting to do for a long time but the paranoia built into me as an American has me thinking I'd be put on a list for scanning infrastructure
@wilkinson.graphics @Peperlito all I see is a dog chasing a turtle