Crawl, walk, run; the making of a Diver Propulsion Vehicle (DPV) mounted photogrammetry sled

Author: Roger Lacasse

Photogrammetry is a powerful technique that allows the creation of detailed 3D models from photographs. The number of pictures required depends on the size and complexity of the target. Underwater, the number of required pictures will also be influenced by the distance the target at which can be seen clearly, as well as the amount of ambient light available. The number of pictures will vary from a few dozen for small objects, to a few hundred for a power boat size wreck and a few thousands for large wrecks.

Naturally, when starting in photogrammetry, divers typically use a hand-held camera or video system and swim around their chosen target to collect data. This is clearly the most affordable way to create models. However, it quickly becomes very difficult to work if the wreck site is in an area with current. Additionally, having to swim around the site quickly creates gas supply and decompression related limits on the size and depth of targets that can realistically be modelled. Data from multiple dives can be combined, but this comes at additional cost and time. These become critical limitations on a deep and large site like that of the HMCS Canada.

Insert Image 1 (Hand-held-camera.jpg): diver scanning a wreck using a hand-held camera system. Photo: Roger Lacasse.

I’ve been looking for a way to efficiently perform photogrammetry scans of deep and large targets for two and half years. Since no commercially available option exists, I endeavoured to develop a solid and stable Diver Propulsion Vehicle (DPV) mounted platform to hold lights and cameras.

From the start, it was clear that such a system needed to be neutral and trimmed. Ideally, it would be suitable for long distance travel too. A first DPV nose mounted photogrammetry platform was developed with inspiration from systems designed by Holger Buss and Brian Bosveld; and assembled with help from Richard Larocque.

The nose mounted rig was based on a 2 m / 6.5’ long collapsible bar that pivoted on a centre axis. The system was mounted on the DPV nose using a flexible-rubber pipe coupler secured with hose clamps. Three GoPro cameras and four lights were mounted along the length of the bar. The buoyancy was obtained by filling the centre mount with closed-cell foam designed for commercial Remotely Operated Vehicles and photography buoyancy floats at the end of each arm. The rig was first tested with diver support from Marc-Antoine Perreault during the early spring of 2024 at Centeen Park in Brockville, Ontario.

Insert Image 2 (First-Sled-2.jpg): nose mounted photogrammetry rig first test run. Photo: Marc-Antoine Perreault.

Although, the rig tests were successful in producing 3D models of large areas at a much faster speed than swimming; it proved difficult to handle underwater. The deployment of the arm sections required careful manipulation before initiating the scan and for storage once the scan completed. The DPV nose mounting made it difficult to handle the DPV in tight turns. The large leaver arms also generated enough torque to spin the bar on the centre pivot post. The lights proved to be too weak for dark conditions. The configuration limited the options for increasing buoyancy and adjust system trim. The mounting rubber pipe reduction also limited the DPV compatibility due to the diameter specific mounting. Moreover, the pipe coupling to the DPV was not the most secure and risked coming undone while travelling at high-speed.

For the second iteration of DPV mounted photogrammetry system, an extended set of requirements was drafted based on the lessons learnt from the first system along with inspiration from the Australian team WreckSploration work. First, the second system iteration still needed to be neutral and easily trimmed. It had to be easy to manipulate on the surface and underwater, requiring minimal manipulations to initiate and terminate a scan. It had to be compatible with multiple DPV’s with an adjustable mounting system that can fit various diameters and dimensions. Following a quick survey of DPV dimensions, maximal compatibility imposed a maximum length of 40 cm / 16” and width 75 cm / 30”. The system needed a way to generate at least 2 kg / 4.5 lb of buoyancy to offset the weight of the system, cameras and lights and provide reserve buoyancy capacity. The floats had to be attached in a way that allowed them to be added and removed at will. It also needed a way to adjust weight positioning to fine-tune trim. Finally, it needed to be easily disassembled for air travel, and fit in a suitcase or a golf bag container.

Insert Image 3 (Second-sled-concept.jpg): photogrammetry sled concept sketch. Author: Roger Lacasse

Those requirements were used to draft a photogrammetry sled concept sketch. A working prototype was assembled in August 2025 with the help of Guy Shockey using off-the-shelf components. The frame was assembled using extruded aluminum channels. The floats were made from 65 cm / 25.5” long 3-inch diameter ABS capped tubing. Mounting points were added for two housed action cameras and four strong video lights with 100 m / 330’ depth ratings. The sled attachment system was based on standard cam tank bands secured to the two frame longitudinal bars. A little over 2 pounds of lead weights was required to achieve neutral buoyancy and trim.

Insert Image 4 (Second-sled-table.jpg): photogrammetry sled prototype. Photo: Roger Lacasse

Insert Image 5 (Second-sled-DPV.jpg): photogrammetry sled prototype mounted on a DPV. Photo: Roger Lacasse

The first test run was performed by myself with diver support provided by Tom Crisp and Jason Cook, and with surface support from Guy Shockey. It lasted 20 minutes and covered the wreck site of the Robert Kerr in the water (13°C) of British Columbia, Canada, and generated 5000 pictures which successfully produced a low-resolution 3D model.

The sled performed better than expected during testing. It proved to be very stable and maneuverable at depth. Because of the achieved trim stability, the sled does require some force to steer, but is very manageable underwater and on the surface. It remains sufficiently maneuverable to be moved around smaller target features. The first test scan was performed at cruising speed while the GoPro cameras were each taking two pictures per second. The overlap and clarity of the pictures are surprisingly good. Since then, additional test runs confirmed that the photogrammetry sled can successfully produce 3D models at velocities even higher than cruising speed and that lighting is sufficient to work in complete darkness. Based on the photogrammetry sled test results, the team is confident that we have an optimized tool to efficiently scan large targets with current and at depths that go well beyond recreational diving limits.

Insert Video (Sled-run.mp4): Guy Shockey driving the photogrammetry sled on a test run. Video: Roger Lacasse