Requesting assistance building DIY ultra long range IR camera

A good few years ago, I worked for a place that did plant safety in the sort of plants where even sparks meant people would probably die. Fire safety was beyond what most of us would even consider "appropriate" but the interesting thing is they didn't use cameras, rather highly specialised IR detectors. 

Cameras are great for thermal imaging and certainly have a place in fire detection  but they're more useful after a fire has started burning. They can be used to see "through" walls and scan an area quickly so that fire control experts can dedicate resources in the right area and this is what we'll look at here:

Curiously for "simple" IR detection you can use a standard digital camera module but these days it means removing the IR filter. (Don't ask, the reasons are ... not what you might imagine.) With the filter in place the camera will still register some near IR but it does mean cranking the gain up as far as you can. What you will need though (regardless of what you're doing) is a "black filter for IR". These filters look so opaque you can just about look at the sun through them but they are transparent to the much longer frequencies of infra-red. You'll probably find instructions on how to butcher a variety of sensors online, particularly for fairly old DSLRs - the sort you can pick up on eBay for a song. Since most of these have a video out, they are quite easy to interface to. You won't need to focus of course, but if you did, lens have correction offset to show where the focal point is. 

So that's one option.

Another is a standard IR module which you can get from China for a few $s - these are made for the Raspberry Pi but I can't speak to their quality. 

The AMG8833 can be found on Ali Express for about £15 - this is a true thermal module with 64 pits and an I2C interface.

If you need a bit more resolution there's the MLX90640 with 768 pits - supposedly sensitive to 1 degree C which is available as a module or just the camera alone. https://www.melexis.com/en/product/mlx90640/far-infrared-thermal-sensor-array

The key to the distance is more about the lens. Just like standard optical lenses (which will focus IR) you can put the sensor behind a lens and that will pull in more data.

Good luck with your project!

@marcdraco Thank you do much for the advice! I think I will go with standard industry units for the stationary long range sensing and then buy some DSLR cameras, butcher them and use them as in flight sensors for real-time monitoring and evaluation. The problem with the sensors you suggested at the end is that they don't have much resolution and in order for my vision model to be most effective I need good resolution so the model has more visual (thermal) data to work with and be more accurate in detection.

On another tangent, does anyone know of a reasonably priced VTOL kit that can accommodate a DSLR weight payload and fly at medium altitudes?

I understand the resolution problem there, this seems to plague IR sensors at any price point. I wonder if you could get around it using an array of sensors and building a map in software. I suspect the issue is to do with the low energy of IR compared to visible light. We tend to think of fire as having a lot of energy because it burns - but that's when we come into direct contact with it. The actual energy radiated by fire over time is small compared to shorter wavelengths, even visible red has more "quanta-fiable" (man that's a terrible joke) energy if we take the integral over time. This is where I hit the limit of my knowledge. 

Traditional camera sensors actually have twice as many green pits as the do red or blue because our eyes have a peak of sensitivity in that area. I've done some IR photography and it requires quite long exposures to get a decent picture. It's such a bind that I often used a technique to fake it, not that that's any use to you.

There certainly are quad and octo-copter drones that can carry a DSLR as a payload, some are even autonomous but I dread to think how much those things would run you. Perhaps someone else has an idea? I'm stuck in the sub-$2000 market with the likes of DJI. 

@marcdraco that is actually a really useful response, I could build an array of them and have the model look at the incoming picture as one then I would have more data to work with or as you said more quanta-fiable data😂 I have reached out to some OEM sources and they're telling me they can build a prototype of a heavy lift drone up to 100kg - way more than I need, so I would use the extra payload capacity to carry more batteries - and then I could hook up a modded DSLR to use for monitoring. Does that sound like a good plan?🤔

Holy moly! 100Kg, that's nuts. Amazing how powerful these things are these days. Over in the UK, we're very limited (it's been a while since I did my exam) but above a certain weight, we're required to take a more advanced pilot's exam. My own licence is for visual rules only (it has to stay in visual range), we can't rely on the on-board camera because it's a point of failure and the RTB might fail too. Plus we can't fly anywhere over people, houses etc. so it's hardly worth owning one. :/ 

But yeah, that does sound like a plan if you can obtain the right model of DSLR, best to check online to see which ones have removable IR filters, but I rather suspect that a sensor array would be a better solution since these are designed to operate in IR vs. just being able to "see" IR.

Even a mobile phone should be able to do this if you put a decent IR filter over the top. Some sellers post a "red" filter but if it passes visible light, that will mess up the exposure. A decent IR filter should be all-but black.