NASA's Dr. Ved Chirayath's Fluid Lensing work as applied to imaging through water from the air is described at length in this ScienceDirect article entitled "Fluid lensing and machine learning for centimeter-resolution airborne assessment of coral reefs in American Samoa."
In 2015, Chirayath's group commissioned a pair of Pumpkin HiPPiECams (renamed "FluidCams" by NASA) capable of 90fps 2kx2k RGB and monochromatic 12-bit image capture, limited only by on-board storage. The two cameras were delivered on time and under budget.
HiPPiECam incorporates an Intel i5 processor running a Pumpkin image capture program on Windows 7, a Pumpkin SupMCU control module, dual SSDs in a RAID0 configuration, and a breakout board with multiple interfaces (Ethernet, USB, eSATA, etc.) to the i5. The imager interfaced to the i5 over USB 3. Pumpkin had a series of Leica 1:2.8/28 ASPH and Elmar-M 1:3.5/135 lenses tested at Olaf Optical Testing and fully characterized for field curvature, etc., so that the complete optical path could be optimized in post-processing.
Since then, Pumpkin has continued to create new high-speed image capture architectures. The latest is nVidia Tegra-based, runs Linux, has direct GPS geotagging support, runs from AC or DC power and supports multiple cameras with GeoTIFF and other supported output formats.