3d mapping software from monocular images


I know by sure that it is possible to build a 3d model of an open area just by taking some pictures of it; I know because I did it:

This is “Barbagallo crater” of Etna volcano, in Italy; I shot ~50 pictures from above, walking around the rim, and the I processed it by Autodesk Remake (now no more free, but ther are also 3d Zephyr and Regard3D still free).

I also know that it is possible to build 3d model of the interior of a room by just a dozen of 360° pictures (I read a paper about this), but I couldn’t find an off-the-shelf SW to do this: I should split by hand each 360 image into 6-10 flat images before making them processing by 3d SW…

The question is:
does it already exist a SW which automatically gets 360° images and turns them into a 3d model?

Second question:
Does it exist a SLAM (Simulteaneous Location And Mapping) SW which can work in realtime on a standalone underwater drone? The algorithm is quite processor-intensive, but it can also be tuned to give coarse results, so maybe somebody implemented it on a Raspberry or something like that?

What I’d like to do is create a 3d model of an underwater cave.

Thanks for your help.


tl;dr: I don’t think it’s feasible with current Trident hardware, I’m a dabbler in the field, so don’t take my word for it. Don’t give up!

A google search for “underwater photogrammetry” brings up a lot of useful results. There are commercial solutions out there but they are for stereo cameras. I didn’t find anything that would work for the single camera Trident. Seems a lot of the software is for high budget projects (archeology, shipwreck reconstruction, etc.) using high end ROVs.

The problem with underwater SLAM is lack of sensors. There’s no GPS, no LIDAR. If a future version of Trident had dual cameras and some sort of sonar array, that would be great. Otherwise we are stuck with the single camera and maybe some dead reckoning based on telemetry data.

I saw some article where they laid out easily recognizable metal plates with a checker pattern around the area of interest to make it easier to link the individual images more easily.

Good luck


Q1 A1: Yes. Free ware is VisualSfM, good tutorials on it at YouTube. A2: Arcsoft, and others exist, top quality but at a price.

Q2 A1: Yes. A2: Not on the Trident at this time. High end ROV’s, AUVs, and hybrids use Sonar/Doppler Velocity Logs for pose and mapping. Very expensive kits.

The SfM that you did with the volcano you can do now with the Trident. It’s a high def camera so SfM will work. However, there are some caveats. You need to use the intrinsic camera parameters for best reconstruction, moving the Trident around can be tricky. You’ll want image overlap for feature matching. Turbid water will make it very difficult if not impossible. I’ve been trying it with Oyster reefs ther ein the Chesapeake. I’ll post more soon.