Also a bullshit conversation about making one at a recent LARP didn't help with the whole "doing the sensible thing" thing.
Unlike so many bits of tech we take for granted as easy in the world of supercomputer smartphones and big data this is still the stuff of movies and expensive law enforcement hardware.
Far infra-red doesn't pass through glass for a start and lenses that can focus it have to be made of things like silicon or germanium.
Then it doesn't work with the usual CCD/CMOS sensors found in digital cameras. There are similar solid state devices using different tech around but they haven't had the development work that devices for visible light have.
A bit of a read around the subject shows that Omron do a very affordable thermal sensor, the D6T-44L. This is quite easy to talk to over I2C and comes in at about £37 by the time you've bought the tiny connectors it uses. Great stuff, but sadly it only has a resolution of 16 pixels. Not 16 megapixels, 16 pixels. It also only does about 4FPS. Still it seemed worth a punt as a first go with this kind of tech as it's quite manageable.
Here I've taken the readings from the sensor and used them to make a pretty responsive display using ye olde MAX7219 and an 8x8 LED matrix. As it already involves some upscaling I've used each block of four pixels to do very basic dithering that represents different temperature readings and made it auto-range.
This doesn't result in desperately sophisticated imagery but I'm happy with how it's come out for an afternoon's faffing around.
Once I connect the sensor up to something more powerful (probably an RPi) I'll try combining it with some conventional video to do a pansharpened image. If this works OK I may end up buying a more expensive sensor like a FLIR Lepton, as I suspect this won't really be usable for much beyond a curiosity.