3D Photogrammetry

Especially for avid wreck and cave divers, creating 3D models using photogrammetry presents a challenge and great satisfaction when it finally succeeds. Camera handling and lighting are particularly important during shooting to achieve good results. The object of interest is captured from all sides with as many photos as possible in series or in a video. It is particularly difficult when it is dark and visibility is poor. We tested this on a small boat wreck in Lake Edersee; visibility was rather modest at about 2 meters:

For better basic illumination, we used a very bright underwater video light.

The team from the first exercise: Sebastian, Marek, Tobi and Stefan

During the subsequent processing, it became apparent that there were some areas with insufficient data:

For example, a part of the bow was missing that was not well lit, and the figure lying in the boat lacked detail.

We repeated the experiment on a sunken car in Messinghausen (See im Berg). Here, almost no additional light was needed, and visibility was about 15 meters:

The model has already improved.

Using suitable 3D software, I use Metashape 2.0 Pro from Agisoft, a point cloud is first calculated from the existing material:

This can then be cleaned up to eliminate superfluous points or errors, which involves quite a bit of manual work. The computation time for this model was about 2 days. However, it's worth sticking with it and experimenting; 3 years ago, my attempts looked like this:

If you are interested in this special procedure, I can recommend the IANTD Diving Skills for Photogrammetry course. Here is the course info