More than any other photogrammetry project, getting a complete and good model of the stern section of the UB-88 has vexed me.
It has been very frustrating and my first attempt failed. Below is a screenshot of the model I made in November 2020 that was only able to resolve the starboard side:
You can get a sense of the structure for sure and can see the single rear torpedo tube and the ghost net. But, it was really not what I wanted.
A year later and a few dives more on the UB88, I have a95+% complete model of the UB-88 at this point — minus the very tippy-top stern end with the stern torpedo tube. With great hope, I went back for a solo dive a few weeks ago to take another pass at building a model and finishing that last 5% of the model with accuracy.
Once again, I was denied.
This is the story of the dive and what I have learned.
My plan was to drop on the down line, orient myself, and then start taking pictures where my previous model left off to capture the very end of the torpedo tube and the ghost net.
Below are screen captures of the model of the port and starboard side of the wreck that show the stern section minus the very end of the torpedo tube. This is the current state of the rear part of the model.
The most critical part of this plan (and the part that ultimately caused the failure) was that I would go down one side of the stern, rotate around the end of the torpedo tube, go back up the other side and the reverse course to get more pictures.
That was the plan and that is exactly what I executed.
The execution was, quite honestly, flawless.
The results were not.
The down line was placed perfectly and the visibility was pretty good. I was on the port side of the wreck so I set myself up and started taking pictures and headed towards the end of the torpedo tube. I rotated around and then went up the starboard side. I reversed course and and took pictures going back and around and then had some time so I took some natural light photos to build a new mosaic from the top down (but that is a story for another time).
I got home and started stitching together the photos and then ended up in despair and I got “ghosted images” similar to what I had the first time I tried to build a model of the very tip of the stern. Ugggghhhh.
Here are some screenshots showing the results of the alignment and you can see where the pictures were taken from (according to the calculations from the Metashape software):
Yeah, it is a hot mess.
I spent so much time trying to figuring out how to make this all work. I tried to align pictures from my dive a year ago with the new pictures but that was another complete failure. I built at least 5 test models to see if I could get things to work.
One of the problems is that the ghost nets at the stern “move” as currents move them around. Here are two pictures from approximately the same place on two different dives:
Not even considering the color differences (I can fix that) but note how the ghost net has moved to cover the end of the torpedo tube. There is just no way that Metashape is going to be able to align photos when that happens.
Lessons Learned / New Approach
So, what happened?
I now know the problem. By going down one side and then pivoting around the end of the stern tube (with a net that isn’t stable and is very hard for software to correctly align no matter what) is just a recipe for problems.
Even if there is just a very small mis-alignment on that very tip of the torpedo tube, then it “magnifies” for other photos that align and hence there are “ghosted” images.
So, the correct approach for this specific part of the wreck, would be to take pictures on the port and starboard side going towards the tip but getting references from either side as you are moving to the stern. It is NOT a good idea to go around the tip as I have done in the past.
Lesson learned (for sure).
I have an almost complete model and I have another dive planned on the UB-88 for mid-November. I will probably publish the full model before that dive. I’d like to get a really good version of the stern torpedo tube into the model but I’m honestly not sure that is even possible with the level of detail I have on the rest of the wreck.
So, for now, here are some pictures of the stern torpedo tube and stay tuned for a post that is a recap of the entire process and (finally) a complete model.
The stern torpedo tube really is a great feature of the wreck.
2 thoughts on “UB88 Photogrammetry Project — Stern Take 2 (San Pedro — 190 fsw)”