Intro
This was one of my “to-do” list projects for 2024. I originally dove the New Saturnia with Tyler and Lora on the Marissa back in October 2020 and we had amazing visibility at a site known for hit-and-miss visibility.
In Aug 2023, Steve Lawson & Chris Gilmartin invited me to go dive the New Saturnia and I jumped at the chance. The location of this wreck is not published and Steve was the person who helped identify the wreck when it was found. We had reasonably good visibility and I got photos for the “back half” of the wreck and vowed to get back to finish the model at some point.
Steve invited me back out this past weekend (Jun 2023) so I figured I would take the opportunity to attempt to finish the project. We got on-site and I was the first down the line with a plan for one long dive. The visibility was 5 feet and very poor. I took a ton of photos but didn’t have a good feeling that I would be able to stitch them together — never mind trying to line it up with the previous dive.
I returned from the dive and started working on the project. It took me a lot of post-processing work, but I was able to come up with a workable model of the wreck. Note that this wreck is mainly a debris field with the massive engine sitting up near the front of the wreck under where the pilothouse would be.
I have included screenshots and links to the model below along with a detailed discussion about the different attempts to build a model. It is probably more than most people want which is is why I placed it after the model — however, it will give some idea of the challenges involved in doing photogrammetry in Southern California in poor conditions on a debris field.
Photogrammetry Model
Below are screenshots and links to the on-line model for both mobile and desktop computers. Obviously, the wreck is a debris field which complicates things quite a bit when building a model and it doesn’t look like a “shipwreck” to most people. I have also include a screenshot that identifies key parts of the wreck and a photo of the original vessel.
Below is an overview of the wreck site with some key features noted.
Here are links to the on-line “live” 3D models for both desktop and mobile:
Desktop (higher resolution)
Mobile (lower resolution)
Below are some detailed screencaptures.
Detailed Process
This section might be of interest to those building photogrammetry models of debris field under varying conditions or possibly for those who just want to learn more of the details of the process to build a model under those circumstances.
Dive 1
The first dive for this project was on 19 Aug 2023 and I took 1,284 pictures and covered the stern area leading up to the engine. I had a 100 minute dive with a bottom time of about 40-45 minutes and did about 50+ minutes of deco. The visibility was decent (on this wreck, that means 10-15 feet) but it is still hazy and any small wrong kick will stir up the soft sand and silt out the wreck.
This is a picture (straight out of the camera) of the tank structure at the rear of the debris field on the starboard side (top right of the model posted below):
You can obviously see that the visibility isn’t great, but it is not too bad (everything is relative).
There are two really tough things to deal with on this type of model in these conditions: (1) the visibility is limited so keeping track of where you are and what you have photographed is tough and (2) because it is a debris field and doesn’t have obvious signposts and structures, it is even more complex.
Those two factors combined definitely make things very hard.
After Dive 1, I had the following model built:
Not bad for a days work in “so-so” conditions but I knew I had a lot of work left to do. The most identifiable feature of the wreck site is the long drive shaft running the length of the wreck. I kept that in mind for when I returned.
Dive 2
In preparation for the second dive, I had printed out and laminated the results of the model after the first dive (people often ask if I take down additional information for multi-dive models — yep, when I can).
The second dive had to wait some time to get completed. We had a bad winter and Steve had a few issues to resolve on the boat. We went out 10 months later on 9 June 2024 and I had just about the same exact dive. It was a 96 minute runtime with the same 40-45 minute bottom time followed by some deco.
On this dive, the visibility was pretty bad. I ran a line from the anchor and it took me ~ 5 minutes to find the wreck. When I did, I tied in the line and put on a strobe so I could find my line tied to the wreck. I couldn’t even tell where on the wreck I had tied into. Even if the visibility hit zero, I could navigate around the debris until I saw the strobe and the follow it back. Yes – the visibility was pretty bad.
These conditions make photogrammetry very tough. Photos need to have a LOT of overlap which means more photos and getting closer to the wreck so the camera can even focus. In fact, I had to crank up the lights, open up the aperture and slow down the shutter speed so the camera would even focus. It wasn’t very much fun. Below is a representative photo:
I took 1,311 photos on the dive.
The first step was to see if I could get JUST the photos from Dive 2 aligned before trying to combine them with Dive 1. I ran a quick alignment on the photos straight out of the camera on Low quality and got something unrecognizable:
That was obviously not going to work so I ran a medium alignment process and had much better luck:
At this point, at least the photos from Dive 2 would align with themselves and you can see how I used the long drive shaft as an “anchor” to tie the two models together. I was hoping that even though almost a year had passed, there would be enough points in common to build a model.
DIve 1 + Dive 2
My first attempt with all photos unaligned worked with Medium alignment. I think I might have tried adding Dive 2 to the photos already aligned from Dive 1 but I’m not sure. In any event, I had 2,587 photos align between the two dives. I build a Medium and then a High quality model as shown below:
This version was “okay” but there is something obvious — the photos on the left (bow) part of the model are definitely exposed more than the right and they aren’t as “sharp.” This is due to the increase in light power I needed to use to get the camera to focus along with probably the particulate matter in the water.
I then used Lightroom to do some post-processing on the Dive 2 photos. I reduced exposure, increased Contrast, Dehaze, and also increased Contrast. I didn’t do anything special with white balance or any other changes. After building another model, I had a better looking “front” (or bow) but now it was darker than the rear. I had overdone it!
You can actually see a little bit of the shadow of the darker photos along the drive shaft. By default, the software builds a “mosaic” of all photos for a specific point.
Now I had a choice to make and decided to do some post-processing on Dive 1 photos to see if I could get everything to match up.
The “final” version is a bit dark, but not too bad. I ended up building a High resolution version and went with it.
Next Steps
I don’t have any immediate plans. If I do go back, there are some areas that I would like to fill in for completeness, but I don’t think they are really critical. See the circled areas in the screen capture below. They are mainly around the front that was so silty and low visibility that it was hard to get some depth for alignment. There is a part of the pilot house that isn’t complete, the area in front of the flywheel is blank and, from memory, there is more debris out near the very bow area.
I would also like the amazing big engine to be a bit sharper, but it isn’t critical.
Thanks
A big thanks to Steve Lawson & Chris Gilmartin for the invites to go dive this wreck. As mentioned, it is a rare opportunity to dive it and building the model has been on my pending list for a while.
Also thanks to Tyler Stalter who helped with some research and who found it the first time we dove it together.
