No, sorry to disillusion you all, but the optics won’t work.
A real thermal imager (like the one you pointed to) typically functions by measuring in the 7-14 micron wavelength range of IR. The imager simply measures the amount of energy collected within that band; its sensor like an array of tiny energy-collecting buckets, each “bucket” energy level coded to a pixel’s color range. The 7-14 micron band was chosen because there happens to be an “atmospheric window” (through air) in that part of the electromagnetic spectrum such that those particular wavelengths can travel without being absorbed. That way distance doesn’t affect the measured energy.
Turns out that water is almost completely opaque at those wavelengths (it might as well be milk). Even with direct submersion, at best all that you’ll measure is the water touching your (IR transparent) lens.
Parenthetically, most lenses (glass/plastic) are not transparent to those wavelengths of IR. If you’re trying to look through glass/plastic, you’ll mostly measure the glass/plastic. Since you’d of course have to house the camera behind a thick pressure window/dome, the thickness guarantees your IR camera won’t even see the wet side of the window.