We recently re-scanned the 12m shelf at Vobster Quay, inlcuding the Hawker Siddeley HS-748 aircraft.
We used the DeepWater Exploration Inc. explore3D stereo camera system - with vSLAM - and lighting from the DWE aquaLUX Max lights. This system, mounted in our bespoke payload sled, attached to our Blue Robotics BlueROV2 ‘Rachael’. Across 2 dives totalling 133mins, we scanned the area initially nadir, then scanned oblique on the second dive. The 3d mesh was processed in Agisoft Metashape from 40k images. Metashape initially meshed at over 29m faces, but we had to decimate to 6m faces for publication. Texture is single 8k.
Are there any particular challenges you’ve run into with this approach, or is it pretty straightforward with the hardware and software you listed?
You mentioned this is a re-scan - I’m curious if there are significant differences from the previous one, both in terms of the scanning process and quality, and notable differences in the scene itself (if it’s changed in a noticeable way). Not sure if you’ve tried aligning the scans and seeing what’s moved, assuming the previous one was at a quality level where that’s even a possible comparison.
@EliotBR - the first time I scanned this aircraft, was with a DSLR as a diver (swimming). I then refined the process using a scooter (DPV). Since moving from real diving to diving by proxy, my first scan of the nose section I did with the standard BR Low Light camera, I then created a basic GoPro stereo rig, followed by a slightly more in-depth rig. Most of the GoPro phase of the story can be found in this thread.
Since having the explore3D, I have scanned the nose section only, on this occasion I decided to push the boat out somewhat.
For reference comparison, there are other companies using the same aircraft to showcase their SLAM collected datasets.
Funny you should say that, just after this current scan, the Nautical Archaeology Society’s training area (the pallets, anchors and small boat at one end of the model) started being moved, as that end of the area is having a few modifications. Once the modifications are complete, there will be a re-scan and a ‘spot the difference’.
Nice! I see that as likely one of the most valuable aspects of detailed scanning / modelling, for scientists and marine asset managers alike. The built environment has BIM, which is gradually becoming more widely used, but many things were made/established before such systems were in place, and it doesn’t apply at all for natural locations and wreckages.
I’ve been ‘dreaming’ of time-tracked spatial model deltas for close to a decade now, and I’m really excited to consider how the ‘digital twin’ space and capabilities will progress from here!
One very cool option would be spatially-consistent Gaussians, which could represent the different fragments of a space, including how it moves and deforms over time, while also being directly renderable using Gaussian splatting…
