Sub Tower Photogrammetry Model & Videos (Catalina — 205 fsw)

I published the story and photos for the “Sub Tower” dive a little over a week ago.

I just published two GoPro videos of the dive along with a photogrammetry model of the docking station portion of the structure:

Sketchfab Photogrammetry Model:

Vimeo GoPro footage

Note that the below footage is just “raw” video without any cool music or editing. I shoot the video just for documentation and review. so don’t expect anything fancy. 🙂

Model Screenshots

Here are some screenshot captures of the current model:

Background

I’ve been spending a little time working on a photogrammetry model based on the GoPro footage I took while shooting stills. The Metashape software can “chunk up” a video into stills for processing. I find that the models don’t come out as well as using stills and have discussed that before.

Challenges

I had another interesting experience on this project. Basically, the video starts at this docking ring, then follows the tower up one side, pauses the top, and then follows the tower down the other side back to the docking ring.

The problem I had is that the model didn’t line up correctly. The smallest error in alignment can “throw off” the model when taken across longer distances. It must have had a small alignment error at the top of the tower which then compounded as it went down resulting in two docking stations that are mis-aligned:

This could also be due to the fact that Metashape doesn’t move EXIF data from video to stills; therefore, the software doesn’t have any way to calibrate the photos. I’ve worked with this in the past when creating a model of the TBM Avenger from the video that Ben Lair shot.

Follow-Up

If I have some time someday, I’d like to build a complete model with correct alignment. I see two options:

1. I could rewrite the EXIF tags and then re-align the photos and see what happens. I might try that at some point.
2. I also came up with another possible solution: start with photos of the docking station from both videos to ensure that those align properly and “trace back up” the tower from both videos on either side and slowly build back up to the peak.