Background
The Missile Tower is an artificial reef created from an old Trident missile testing tower. It is very close to the US/Mexico border and sits in about 160 feet of water. It is a popular dive site for technical divers and is commonly used for training classes since it is big target and it isn’t too deep so you don’t rack up too much decompression time too quickly.
I’ve written about the Missile Tower in the past and have dove the site a number of times. I’ve always wanted to build a photogrammetry model of the wreck. This post will cover a little bit of the history of the wreck, my previous attempt to build a model, and the results of my most recent attempt (which was successful)!
Trident Missile Tower
Below is a map from a California Artificial Reefs document which shows the location of the site. I believe the artificial reef was created in 1992 or 1993. There are a lot of other beams and rock / rubble in the surrounding areas.
Below is a detailed map of the general location. According to the document, the Missile Tower is at 32° 32′ 29.7″ North, 117° 14′ 47.4″ West.
The picture below is of a Trident Missile Tower from the Divebums website. You will notice some similarities to the photogrammetry model below. The tower is on the side and is about 60 feet long, 25 feet wide and 25 feet high so the top of the structure is at about 135 feet. As you will see, the wreck is covered in white metridiums (which complicates building a photogrammetry model).
For readers that aren’t aware, Trident missiles are used on submarines and are launched from under the surface of the ocean straight up into the air. The amount of engineering that goes into building a system like the Trident is just amazing.
First Attempt
In June 2021, I decided to see if I could start to build a model. I was still early in my photogrammetry learning process at that point. I think I was just on a “regular” dive and didn’t have a ton of time but thought I would at least see if I could start a model or see how far I could get. I took 464 photos but could only get about 200 of them to align reliably. I did a few different test builds, two of which are below.
Build 1 – 194 Photos
This was the most accurate build but I could only get about 50% of the photos to align.
Build 2 – 300 Photos
In this case, I was able to get 300 photos to “align;” however, they weren’t really aligned correctly as can be seen in the screen capture below. You can see how there is a duplicated end where the missile tube is. I think this happened when it misaligned around the boom / steel structure that sticks out towards the sand.
I didn’t spend more time on the project at that point because I knew I would have to get better at photogrammetry, spend more time thinking about how to attack the problem given the type of structure and spend a dedicated trip on the project.
The Dive
I had been doing some diving in the general area recently and the visibility had been very good so I wanted to get back out as soon as possible to make an attempt at building a model. I recruited Drew Wilson for the project and we booked a trip on the Marissa with Chris & Lora. Drew and I do a lot of the projects together. I swim around with the high power Keldan lights and drew acts as a safety diver and also takes natural light photos with me providing light & color.
Chris & Lora nailed the downline placement and we were only a couple feet from the tower. I could start to see the white metridiums from about 80 feet deep which indicated that the visibility would indeed be great. Drew & I spent about 45 minutes on the structure with me doing quite a few laps around it to make sure I got enough photos and enough overlap. There are a lot of nooks and crannies to cover.
As I predicted, it was going to be a challenging project. We had some current deep which meant I got pushed into the wreck in one direction and pushed away in another. Also, the two biggest challenges were (1) the massive amount of fish on the wreck which wreak havoc when trying to focus on structure and (2) the large number of metridiums which can sway slightly in the current and the coral on the top of the structure that would also move with the current.
I took about 1,500 photos and hoped for the best.
Drew got some really nice natural light photos including an interesting perspective from inside the missile tube looking out as I swim by.
The Model
I thought about trying to use photos from the original project; however, the metridiums are not static and I’m pretty sure that alignment would never work accurately. This project needed to be completed in a single dive (or possibly two dives in the same day)
I made probably 7-8 laps around the structure. This was definitely a challenging model due to quite a few factors:
- The overall size and complexity of the structure
- The number of fish that were “distracting” to the auto-focus and covered structure
- The white metridiums which sometimes moved and definitely covered a good portion of the structure.
- There was a bit of a current around the structure which would push me into the wreck in one direction and away from the wreck in the other.
Below are some photos that show the amount of fish and how much the metridiums cover part of the structure. Trying to get the camera to focus on structure and not moving fish and to find spots that don’t have anything moving is quite complex!
Given all of those factors, I’m pretty happy with the results. One thing that caught my eye and looks askance is the cylinders on the top of the structure look askewed from the rest of the structure. They point in a slightly different direction. I’m not sure if that is just how the wreck is or if it is the result of a small mis-alignment that got magnified.
Build Stats
For those interested in the details, the model has 1,387 photos aligned which were taken with a Sony a6400 and a WWL-C wet lens (I like to take my smaller camera when I can for these larger projects). I use a Mac Studio computer with a M2 Max CPU which has 12 cores and 38 GPUS and 64GB of memory (i.e., it is pretty powerful). The final model:
- Built with “Ultra High” Quality and a “High” number of Face Count
- Took about 20 minutes to align the photos with pre-select enabled (after multiple different tires in different combinations)
- The Depth Maps with Mild filtering took 5.75 hours
- The Mesh took 45 minutes to build
- The 4k x 4k texture took 3.5 hours to build
- The Ultra High model has 74M faces
Screen Captures
I really like the look of the white metridiums on this model. It gives a good sense of what the structure looks like.
Desktop Version
Mobile Version
References
Thanks To
- Drew for working on the project with me, being a safety diver, and taking some great photos during the process
- Lora & Chris at Marissa Charters. We had a really great day on the water with them (and Scout & Captain, of course!) and they nailed the downline!
