I don’t remember the first time I saw a photogrammetric model of a wreck but I’ve always been impressed by the technology.
In case you don’t know what photogrammetry is, I’ll give you my simple definition: combine multiple 2D photo images into a 3D model. There is a lot more to it, but that will suffice for now.
One of the big advantages is the ability to study a wreck (or cave or anything else) from the comfort of your desk. When you are diving (especially at deeper depths), you are concentrating on a LOT of different tasks and don’t always see everything. Studying pictures or movies after a dive can help discover what you missed, but it also helps to have a 3D model that you can zoom in and out, measure distances, change perspective, etc.
The most famous underwater photogrammetry project is of the Thistlegorm which is one of the top wreck sites in the world (I have yet to dive it). If you want a mind-blowing experience, go check this out: https://thethistlegormproject.com/3d-photogrammetry-survey/.
I’ve seen a couple videos recently on Wetpixel about photogrammetry and with Covid still limiting my ability to travel, I figured I would take some time to learn photogrammetry. I studied videos, read manuals, and then decided to start testing things out.
Based on a lot of studying and some research, I decided to use Metasoft software from Agisoft.
Project 1 – Statue
For my first project, I used some test photos available on the Agisoft website. The photos are there to allow users to get familiar with the workflow of uploading photos, aligning the photos, building the mesh, etc. Here are some screenshots of my first model:
Project 2 – Outdoor fireplace
For my second project, I moved on to taking pictures myself of a fire feature outside in our yard. I figured that I would start easy and see how things went. It actually turned out pretty well and I’ve included some screenshots of the model below:
Project 3 – The Unknown Sailboat
This was my first underwater test project. In case you didn’t see the post, we dived a sailboat a week ago that had (as far as we know) never been dived before. I figured it would be a good test subject since it is small-ish and the structure is still intact.
It turned out “okay.” Not too bad for a first try — especially given that the wreck was at 200 feet deep.
I put all the photos into the Metasoft software and … thanks for playing but try again later!
Unfortunately, the software couldn’t align all of the photos so I had to break it into three “chunks” and then I tried to align and merge the chunks but that wouldn’t work either. It appears as though I was missing a few pictures at key angles.
Here are some key lessons I learned while building up my photogrammetry skills:
- Don’t mix the use of lights and natural light photos – I’m not 100% sure, but I think this makes it harder for the software to stitch the pictures together into the model. It certainly makes the creating the texture model a bit odd.
- Don’t mix lens zoom levels
- Try to maintain a consistent distance from your target
- Make sure to get pictures at key transition points – corners, etc.
Stay tuned for further attempts at building more photogrammetry models and I will post them on-line so people can use them once I build one good enough to do that!