Accuracy Defined?

Underwater GPS meets Photogrammetry

For a while we have been using GPS located cameras to construct highly accurate models of defensive structures known as pillboxes. Fixing a camera position above the waves is straightforward enough and you can build representative models without additional scale bars or measurements if the GPS signal is accurate.

The problems start as soon as you dip beneath the surface. Radio waves – signals from those GPS satellites – do not penetrate water. Knowing where you are with a direct GPS signal when underwater is just not going to happen.

But its something we have wanted to use for a long time. Work with television companies at Alikanas Bay opened up the potential but remained cost prohibitive. The float lead would tangle and restrict depth…the solution was not a good fit for photogrammetry.

For a while we have been working on the Vobster Photogrammetry Project. A small team of like-minded divers working towards mapping Vobster Quay Inland Diving and its submerged attractions. As the project progressed the problems of disorientation, accuracy and inefficient data capture started to cause problems. What we needed was a navigation system.

Enter UWIS

Finnish company UWIS Oy appeared to be making what we needed – GPS based underwater tracking and navigation. Not only did they have some really cool equipment but their CEO Pertti liked the idea of the project and offered to help!

A few days of January were set aside for showcasing and working with the UWIS system. Comprising of three floating GPS antennas, tracking units, a tablet navigator and topside laptop the kit allows both diver and supervisor to know where they are, where they have been, and where they need to go next.

Briefly, its a diver and supervisor dream – particularly when working in low visibility.

The tablet navigation software had been preloaded with the Vobster ortho photo…so we navigated not by marks or points, but by the image derived from the 3D model – this was a very cool bit of kit!

Fit For Photogrammetry?

Would the GPS data gathered work for photogrammetry? Could we reconstruct an accurate model from camera position alone? How accurate would the model be?

We will confess to being slightly sceptical at this point. Could an underwater GPS signal locate the cameras accurately enough to scale from? There was only one thing to do; test the theory.

The plan was to scan the rusting remains of the AFV432, a small armoured personnel carrier sitting in the lake. Although we have scanned it before target was chosen for several reasons; it was small enough to complete on a single dive, was complex and detailed, and there were preserved examples to take direct measurements from and compare accuracy.

The GPS track was downloaded (see above video) and processed into Metashape Pro camera coordinates. The images were processed, aligned , refined and locations set using the camera positions. The model was set to process and the results were

Accuracy from GPS?

This was the first test of a new process, using GPS to build a scaled model. Early indications are very positive indeed..the variance was very low when check bars were added; sample measurements of 250mm measured 260mm – a variance of 4%.

All the pointers suggest we are onto a completely new way of working with underwater photogrammetry. Using just camera position to create and scale reduces tasks and work that needs to be done underwater. We are back at Vobster Quay this week for more testing and verification, not least of which will be checking variance over a much larger area.

Watch this space!