Scanning an object submerged in water?

[einatlev]

Hello!
We would like to scan an object that is submerged in a tank of transparent fluid (e.g., water), and cannot be taken out of the fluid. Would PhotoScan work as normal in this case? Any special procedures we should employ?
Thanks!
Einat

[bigben]

Yes. Is the camera in the tank as well or outside.  If outside, refraction will cause problems if you're trying to go all around the model.

[azmann]

We've had good success shooting through water using SfM to get river bathymetry, so long as the water is clear and not so deep that the bottom isn't well lit. Accuracy does suffer a bit, however.

[einatlev]

Thank you to both repliers! Encouraging.

We are planning an experiment looking at analogue "lava flows" made of wax that will be extruded into a tank filled with sucrose solution. Cameras will be n the outside.

Azmann -- I would love to hear more about what you did for the river bathymetry!

BigBen -- is there a way to go about correcting for the refraction some how? I would really like to avoid putting the cameras inside the tank... it's going to be gooey and sticky.

Thanks!

[bigben]

I've been helping one of our museums experimenting with tissue specimens in formalin.  Contrast is bad with CT so they're trying photogrammetry.  Through a single face of the tank isn't too bad, but a full 360° is proving problematic.  You could try a separate mesh through each side of the tank and then merge the meshes.

They've only just started experimenting so if we get anything working I'll post back.

[James]

I'm no optics expert, as you will see, but i would think refraction would be a big problem whether you keep perpendicular to the tank or not.

I attach a sketch of a possible solution, which won't completely solve the problem as the sucrose solution presumably has a different refractive index to glass (i don't know how much different - it might be negligible?), but should at least counteract the effect of refraction when passing through the air to 'non air' interface.

A concave-plano lens placed on the tank glass surface, and moved with the camera, where the radius of the concave curve is equal to the distance from the lens to the camera's lens nodal point, which is perfectly centered at that point, should ensure that all rays passing through the concave plano lens toward the camera's standard lens exit more or less normal to the concave curve and therefore do not appreciably refract.

This may be practically very unlikely if not impossible but someone might like to have a go!

*edit - but now after looking at my own sketch i can see that in the example where the rays refract and there is no special lens, it is just the same as standing further back and shooting with a longer focal length, so its probably fine so long as you keep perpendicular! do you even have to keep perpendicular?

[einatlev]

Thanks everybody! Definitely lots to consider.