LWIR, MWIR & SWIR side by side

[Ultrapurple]

Those are most remarkable! Thank you for taking the time to do the experiment. I must try something similar when I next get the chance to get the cameras out.

[Ultrapurple]

This has inspired me to do some experiments combining X-Rays of different energies in an attempt to get 'colour'. Fuller details are here, but here's a quick sample:

[Lambda]

Thank you a lot Ultrapurple, Vipitis, for sharing your experimentations about this fascinating subject of multi-spectral monitoring.
I am also currently in this thematic for my own little experiments. Your teaching is very helpful!

One delicate point, in my humble opinion, is to finally find a "color code", a manner to obtain an image wherein false colors are well selected, in order to avoid a final unclear or unreadable image, a bit too "psychidelic".... Too much informations killing the information i a sense.


Thank you again for your work!

Best regards.

Stéphane

[Ultrapurple]

One delicate point, in my humble opinion, is to finally find a "color code", a manner to obtain an image wherein false colors are well selected, in order to avoid a final unclear or unreadable image, a bit too "psychidelic".... Too much informations killing the information i a sense.

I agree with you.

At this early stage I am only trying to get 'something', without worrying too much about 'what'. Once (if?) I've found a workflow that can produce useful data, I can then go on to working out how to present it.

A great deal of work has been done by the ultraviolet imaging community on sorting out palettes that best show colours from UV; our cases aren't much different. I invite you to join the friendly bunch there - or at least read some of the posts.

There has, of course, also been a lot of discussion on palettes in this forum.

My own preference, for single-band (usually LWIR) imaging, is to stick to monochrome. There are times when colour can be useful of course, but I shoot far more black and white IR than colourised shots.

[coppercone2]

it had something with diamond computers but i think they used satellites and thermal imaging to find the areas where the jungle regrew after being cut down by settlers to make the mining towns because it grows back different (second stage jungle), and despite the long abandoned nature of it the difference still remained, which is how they found whatever they were doing in that book

[bap2703]

One delicate point, in my humble opinion, is to finally find a "color code", a manner to obtain an image wherein false colors are well selected, in order to avoid a final unclear or unreadable image, a bit too "psychidelic".... Too much informations killing the information i a sense.

Are you just trying to find a combination of channels that produces eye-analyzable images when just mixed together, or do you try to get a pleasing results for already analyzed data ?
Usually you don't just mix multi/hyper-spectral data and just hope to see something, you first segment the data either from prior knowledge of the physical properties of what you're looking, or through automatic segmentation to group spectral features together.

it had something with diamond computers but i think they used satellites and thermal imaging to find the areas where the jungle regrew after being cut down by settlers to make the mining towns because it grows back different (second stage jungle), and despite the long abandoned nature of it the difference still remained, which is how they found whatever they were doing in that book

Vegetation analysis screams for NIR and SWIR.
You might be interested by that : https://www.usgs.gov/core-science-systems/nli/landsat/landsat-surface-reflectance-derived-spectral-indices?qt-science_support_page_related_con=0#qt-science_support_page_related_con

[Ultrapurple]

Are you just trying to find a combination of channels that produces eye-analyzable images when just mixed together, or do you try to get a pleasing results for already analyzed data?

Usually you don't just mix multi/hyper-spectral data and just hope to see something, you first segment the data either from prior knowledge of the physical properties of what you're looking, or through automatic segmentation to group spectral features together.

Agreed that in professional applications there are well-researched methods of combining data to achieve specific, clearly-defined results, as described in the papers you link.

From my purely hobbyist perspective, the fun comes from making 'stuff' happen. Unfortunately I don't have the spare cash to buy even a used Landsat satellite, nor even to hire an aircraft equipped with multi-spectral cameras. In fact, I am pretty sure that anything remotely approaching a multi-spectral camera would be way outside my league.

Instead, I have fun collecting disparate pieces of equipment, usually of widely differing vintages and capabilities, working out methods by which they can be made to work together, one way or another, then playing with the results to see if I can learn anything or make any pictures that I find interesting or attractive. This near-UV photo of an apple on the tree didn't break any new scientific grounds, but it did teach me - back in 2013 - that I can make a more or less usable image in the near-UV spectrum.




As a result I know that I can get one channel of a multispectral image that way, with the particular field of view offered by that setup (in that instance, a 35mm lens on a 1.6 crop factor DSLR). I knew from other experiments that I can get a good near-IR image using basically the same gear with just a different filter (and a wildly different exposure). But I also know I don't have a filter for that setup that will give me a decent visible light (RGB) image - I'd have to use a different camera body for that. And since making that UV exposure I have acquired a 1.6 crop factor DSLR that is mechanically similar-enough to the IS Pro used here that I could also make a colour photo, of comparable resolution - thus giving me five channels to play with.

More recently I've acquired the ability to make megapixel-class images at LWIR and, more recently still, half-megapxel at SWIR and rather low resolution MWIR, meaning five or six data channels. Now it starts to become interesting - it's a massive challenge merely trying to get a similar field of view on those radically different imagers, let alone extract the data and process it into a common size and format that I can then manipulate. But when - if - I ever get there, I expect the sense of achievement will be significant, even if the image is only of an apple, or maybe a plastic figurine clasping a wilting flower.

Yes I'd love to find an eBay auction for a wide band, high resolution imaging spectrometer that could output TIFF files. Preferably with a $10 Buy It Now option. But until that day I plan to keep plugging away with the facilities at my disposal, as and when time permits. It's only about ten years until I can retire...

[Ultrapurple]

This is the only remotely hyperspectral image I've published to date. I made it in 2012.



Red represents NIR (1100...850nm)
Green represents visible (R+G+B)
Blue represents near-UV (380...320nm)

More details here.

[Lambda]

One delicate point, in my humble opinion, is to finally find a "color code", a manner to obtain an image wherein false colors are well selected, in order to avoid a final unclear or unreadable image, a bit too "psychidelic".... Too much informations killing the information i a sense.

Are you just trying to find a combination of channels that produces eye-analyzable images when just mixed together, or do you try to get a pleasing results for already analyzed data ?
Usually you don't just mix multi/hyper-spectral data and just hope to see something, you first segment the data either from prior knowledge of the physical properties of what you're looking, or through automatic segmentation to group spectral features together.

Hello,

My objective is very humble and also in accordance, like Ultrapurple, with a context of Hobbyist having (very, in my case) limited ressources.
I have passion for night vision technology and try also to collect materials (optics, image intensifier tubes, low light sensitive camera,...) and data, literature on this subject.
And i was since years fascinated by the fusion technology or COTI principle which was primary used by militaries and more recently in thermal monitoring for building or machines, PCB..... (cf Fluke products, SEEK,....)

I try to implement and to show to myself that such principle and technology, in 2021, can be reached by a hobbyist. As in may other domains (radio Ham, astronomy, general electronics and informatics...) we have now, in 2021, access to a plethore of materials which would have been under the monopole, 30 years ago, of gov agencies/institutioms, univ. lab, R&D department in some companies.... The rate of technology transfer from this confidential or restricted world to our hobbyist, civilian world, this last decade (at least) is amazing, thanks to the open source philosophy, fablab, fundamental data available in the net, archives....

And night vision multi-spectral system belongs to this kind of fascinating playground for me...

https://www.google.com/search?q=night+vision+fusion+image&tbm=isch&ved=2ahUKEwjJkoPinanxAhXhyLsIHdvmAVsQ2-cCegQIABAA&oq=night+vision+fusion+image&gs_lcp=CgNpbWcQAzoGCAAQBxAeOgQIABATOggIABAHEB4QEzoICAAQBRAeEBNQ26cBWNq3AWDUvgFoAHAAeACAATyIAYMEkgECMTCYAQCgAQGqAQtnd3Mtd2l6LWltZ8ABAQ&sclient=img&ei=3snQYMmrBOGR7_UP282H2AU&bih=722&biw=1536&rlz=1C1GCEB_en__891__891

Other example based on digital low light camera working in Vis/NIr (Sionyx) combined to a thermal system...



etc....

Here is my first modest attempt in this direction...

https://www.eevblog.com/forum/thermal-imaging/attempt-for-a-multi-apectral-night-vision-system/

Nothing new under the Sun, just the desire to understand a certain type of technology and to be able to reproduce, even crudely, such one....
and combining Vis/IR and LWIr domain for compensation intrinseque deficiencies from systems working in either one or another spectral domain is very interesting...

For the interpretation, or readability of the image obtained in real time ("video" system), for the time being i am focused on the idea to use one color with associated gradient for one channel (LWIr) and one other color with gradient for the other channel of interest (Vis/Nir). No more in term of colors variation...

I would consider even to keep a simple B&W gradient for one channel and only one color with gradient, and a relative transparency or opacity for one of the two channels for the final video image i would like to see.


Juste some thoughts....

Regards.
Stéphane

[coppercone2]

Can  you use a triple beam splitting setup after a single lens to three different sensors, or do you basically need different optics for the three sensors?

Its interesting if anything because germanium prisms seem interesting to work with. does this kind of stuff exist?

the RF equivalent would be Antenna + triplexer for different bands (with lens being antenna). So you can get a real overlay.

[Bill W]

Doubt Ge is much good at SWIR, and getting good coatings for MW and LW is tricky.

There is also Calcium Flouride - clear to visible and works at LW.

[coppercone2]

that's not a great material for the military but for scientific work it should be OK. I assume its a hydroscopic crystal?

The nice thing is, that is likely a soft and cheap material, however I don't know if its one of those salts that makes HF when it gets wet. Some fluoride salts make HF gas (not F2!) when wet IIRC.

Can these salt crystals be rapidly cut on a lapping plate with diamonds? I guess it depends on MOH scale at this point. Don't know much about machining fused salts... I imagine they are 'crumbly'

[CatalinaWOW]

that's not a great material for the military but for scientific work it should be OK. I assume its a hydroscopic crystal?

The nice thing is, that is likely a soft and cheap material, however I don't know if its one of those salts that makes HF when it gets wet. Some fluoride salts make HF gas (not F2!) when wet IIRC.

Can these salt crystals be rapidly cut on a lapping plate with diamonds? I guess it depends on MOH scale at this point. Don't know much about machining fused salts... I imagine they are 'crumbly'

If you are willing to deal with the drawbacks of a salt, NaCl is as safe as anything gets and good from 0.2 micron to 15 microns.  NaFl has been used in science instruments (mostly during and just after WWII).  Most places went for other materials as they became available.  The relatively soft material was prone to mechanical damage and keeping humidity low enough to avoid moisture problems was an issue.

While these materials could be fabbed in an amateur lab, I think coatings are a problem even for the pros.  So significant reflection losses and possibilities for ghosting.

[Fraser]

ZnS is a broadband material and can even be used in the visible light spectrum. BB AR Coatings might be a challenge though.