Hi I am very interested in this platform to replace my horrible Qysea E-Go. I would like to move into undervater photogrametry and later also underwater mapping and harbour wall inspections. I would love to get some feedback to what you would consider a minimum setup, for an efficient workflow. I have been thinking about the heavy configuration, with the strong thrusters, 4 lights, a cerulean forward scan and maybe a DVL. Or even maybe 2 ceruleans since I read that that also enables Dvl like position hold ?? I would love to hear how you would spec a work ROV ? Also what is the cheapest way to to get some absolute positions. Is it possible to “just” use a DVL, and manually enter the homepoint / the insertion point and have the ROV work with that as a reference point ?
Hi @JakobHK,
Happy to help!
Much of my recent underwater photogrammetry experiments use a BlueBoat with a towfish. This setup covers vastly larger areas than an ROV could. Stay tuned for a guide on this, similar to the Building Subsea Cameras and DART guides!
However, vehicle choice depends on your specific goals:
-
ROVs: Ideal for harbor walls, high-detail inspections, and vertical structures, though less efficient for broad coverage.
-
BlueBoat: Ideal for rapid, large-scale area mapping - especially with sonar (checkout the Omniscan 3D!)
Thrusters: T200 vs. T500 The standard BlueROV2 uses T200 thrusters. While Blue Robotics sells the larger, higher-thrust T500 thrusters, mounting T500s requires building a custom vehicle frame to handle their physical footprint and power demands.
Sonar & Positioning Systems
-
Imaging Sonar (Omniscan FS): This sonar isn’t directly usable in typical photogrammetry workflows because it outputs 2D acoustic imagery rather than 3D point clouds. It is primarily useful as a visual reference when reviewing processed results.
-
DVL & Absolute Positioning: A DVL tracks relative velocity to help the autopilot hold position - this inferred position could be used to scale and locate your captured images. However, you can also scale models using stereo-pair cameras when capturing, scale bars/known object size, or known geo-referenced points. For true absolute positioning fully compatible with BlueOS, the Water Linked UGPS and Cerulean Omnitrack are the most affordable solutions available.
Software
-
Agisoft Metashape Pro (~$3.5k): Straightforward to use with consistently high-quality outputs. Pair it with a powerful workstation (high CPU core count and dedicated GPU) - this is basically the industry standard
-
WebODM: A solid free/open-source option, though it requires more manual tweaking and troubleshooting to achieve comparable results.
-
RealityScan, MipMap, and Postshot are other software tools I’ve tested with a bit - I’m working on a general guide to photogrammetry to compliment the Towfish guide as well!
I hope that’s useful feedback!
Hi @JakobHK,
I have written a few posts about my different photogrammetry setups, which you may find helpful.
The above kit, on its own will supply really good visual quality 3d models, however, they are not scaled or georeferenced.
As with the first link, the second & third links use non-‘survey grade’ hardware, however the ExploreHD cameras can integrate geo reference data from the Mavlink - so IF you have a USBL or DVL with dead reckoning, ‘some’ scale and geroreferencing can be achieved. I just remembered a video I did showing the flight that DWE write about:
The final link is a video, which is a bit of an evolution from the last video link. This one shows the ROV scanning a shipwreck, as well as the topside prep etc. The camera system is a huge step up from the previous evolutionary steps, it is the DWE Explore3D, SVC Pro & Lights. This setup is a calibrated stereo system, which means you have scaling. I have been putting in scale bars to check the scaling, and it is cock-on! The system has SLAM, so you can see what has and has not been scanned in a live view. The system currently does not support geo reference, however I am told that it is on the way.
Now, onto answering your questions…
&
There is no single, simple answer to this - it is also important to understand that you need to consider the means of integrating the georeference data with the photogrammetry data. Absolute accuracy will generally increase as cost increases - but that is not so say always. If you are doing close in, shallow, it is feasible to put some submerged ground control points, that you can survey with survey pole, but also not ideal. If you are working alongside a referenced MBES dataset, it is feasible to merge the datasets and use the MBES referencing, but also may not ideal in many cases. You could use dead reckoning, but this can be very hit and miss, and will hugely depend on the lake/sea bed makeup and bathymetry - another video of when dead reckoning goes well. You could drop shot lines on GPS coordinates, and use those as GCPs. Finally, and for most cases the optimum option, there is acoustic positioning.
I appreciate that the following question was posed before the one I answered above, however I think the answer to this is the most important thing for you to consider.
What do your clients require? Do your clients only want a simple visual representation, do they need scale, do they need geo reference? Your clients requirements are how you spec the ROV.
WOW WOW WOW, you blow my mind. That is awesome information and very very useful for someone trying to leave the Qysea platform and move on :0). I really want to move into this field when my economy allows it, and I would especially like to pair it with a Blueboat later down the road to be able to do seabed scans. I will read through your answer and I will learn a lot I can see, maybe there will even be a followup question. Thx for taking the time, this is truly what makes the WWW fantastic, something thats easy to forget in this day and age, thx :0)