In the ongoing project of building a complete, detailed photogrammetry model of the UB-88, Drew Wilson and I went back with Ray Arntz on the Sundiver for two more dives. On my previous dive, I completed the port and starboard hulls to complete the front half of the model.
My goal on these two dives was to rock on the hull sections and the deck to the aft of the conning tower.
We’ve had a large south swellS and a high-surf advisory earlier in the week for Orange County. Generally, when we get those long, deep swells with long periods, it tends to stir up the deep water which takes some time to settle down. Sure enough, when we got down, the visibility was 10-15 feet at depth (190 feet to the sand). Luckily, Ray had dropped our down line less than 10 feet from the wreck.
Drew had planned to take some pictures to build a mosaic of the top of the sub but given the visibility, I’m not sure how well that will turn out. We’ll play around with the photos to see what we can build.
The visibility doesn’t impact detailed photogrammetry work too much. If you are planning on doing a less-detailed model with fewer photos from farther away, it would presumably impact that.
On the first dive, I spent the dive taking photos while going down the hull and up onto the deck on the port side. I returned back up on the deck towards the conning tower and then ended the dive. On the second dive, we had a bit better visibility and I repeated the process on the starboard side of the wreck.
When I got back, I attempted to do a quick alignment using a “Low” quality match algorithm which didn’t quite work so I had to spend the compute time using a “Medium” quality match and that worked much better. I got a pretty much complete model of both sides of the aft section but without trying to work on the “ghost net” or the very end of the stern torpedo tube given the complexities I had when I tried this in the past.
I took a total of 922 pictures and an amazing 914 aligned on the first pass using Medium quality. I must be getting better and ensuring that there is enough overlap to get a good alignment and also being the sweet spot of too close vs too far away and maintaining that separation.
Below are some screenshots of the sparse cloud and the final textured model. You can see the start of the torpedo tube at the back and a portion of the conning tower at the front. Hopefully I got enough of the conning tower to join the front and back models. That will be the next step.
I think I need one or two more dives. I need to finish up the model at the very end of the torpedo tube and then I’m thinking about having a “model” near the conning tower to get an idea of human size.
I also need to measure the opening in the hull behind the conning tower in order to get an accurate scale model and I have a few other smaller things I want to do.
Then, of course, build a complete model which I’m guessing will take 1.5-2 days of compute time.