Night vision technology vs thermal imaging

[fog37]

TL;DR Summary

understanding night vision technology vs thermal imaging

Hello Everyone,

I have a thermal imager and I am interested in better understanding night vision (NV) technology. Apparently, NV goggles respond to NIR and SW
See this article. NV goggles can see from the near infrared (NIR) to the shortwave infrared (SWIR) portions of the light spectrum.

Do NV goggles work in complete darkness or always need some minimum level of ambient light (like from the moon)?
Why do the NV goggles capture NIR? The human body emits primarily in the far-infrared (FIR) which is detected and imaged by thermal imagers. I guess the human body emits also some NIR and SWIR...

Anyone with some experience with NV goggles?

Thanks for any comment.

 

[Drakkith]

Do NV goggles work in complete darkness or always need some minimum level of ambient light (like from the moon)?

Night vision goggles need an artificial source of illumination if you're trying to generate images in real-time (video cameras, wearable goggles, etc). Still images can be taken with only natural illumination, but they would have to be very long exposures and depend on the availability of said illumination sources, so they would be hard to make when the weather is poor or the Moon is not visible.

Why do the NV goggles capture NIR?

You can make night vision sensors using similar methods and materials to visible light sensors, so they are relatively cheap to make and they have comparable resolution to visible light cameras. You just need an illumination source usually. Many home security systems that use a night vision camera have the illumination built into the camera itself.

Thermal imaging sensors, as far as I remember, are very different from visible and NIR sensors and have much worse resolution. But, they obviously have an advantage that they don't need an illumination source.

The human body emits primarily in the far-infrared (FIR) which is detected and imaged by thermal imagers. I guess the human body emits also some NIR and SWIR...

The human body emits virtually no NIR. It's simply not hot enough.

 

[Rive]

Do NV goggles work in complete darkness or always need some minimum level of ambient light (like from the moon)?

A type of night vision is based on near-IR: this requires (artificial) near IR light illumination. Most surveillance cameras are based on this, since it's cheap.
Cheap, but needs a near IR source: also, while the objects closer are visible, but the background remains dark.
Pretty much like a torchlight with 'invisible' light. Nothing to do with real 'thermal', actually.

A different type of NV is based in light amplification. This does not require any artificial light, since it's just amplifying the (visible) light already present. Though this does not work in complete darkness. On the other hand, this has no distance limits.

Third type of NV is based on actual thermal imaging.

 

[fog37]

Thanks Drakkith and Rive. Let me paraphrase if ok to make sure I get it:

-Thermal imagers: they detected far-infrared (FIR) emitted by objects with a high enough temperature T to emit mostly in the FIR region.

- NV systems:
a) Light amplification based NV systems which don't need any artificial light as they amplify the low level (visible) ambient light. Complete darkness = no light to amplify. The ambient light available at night is provided by the start and the moon. The light reflected by the moon has the same spectrum as sunlight. So no NIR is used in this technology. The images produces by NV system are greenish. No distance limits. These systems also detect the NIR available in the nocturnal light. The nature article I found show the spectra of night and twilight (https://www.nature.com/articles/srep26756.pdf):

I read that there are 4 generations on NVD: at gen 4 NVD (gen 0,1,2,3) that don't need an IR illuminator. This article talks about it in detail: https://www.degruyter.com/document/doi/10.2478/s11772-013-0089-3/pdfb) NV devices that use NIR (700nm -1000nm, correct?): these systems need a source of NIR (IR illuminator) that illuminates the object with NIR (not visible by our eyes). The object then reflects the NIR and the reflected NIR is detected by the onboard NIR sensor. Why NIR? NIR is very close to visible in terms of wavelength when compared to FIR. NIR sources and NIR detector are cheap... I think these are called Gen 0 devices: the unit projects a beam of near-infrared light. Invisible to the naked eye, this beam reflects off objects and bounces back to the lens of the NVD.

Thank again!

 

[Melbourne Guy]

Third type of NV is based on actual thermal imaging.

Sorry, no experience with goggles, but I worked on a machine vision project last year that used an IR camera for crowd analysis, it gave results like @Rive's example photo. I was surprised by the low resolution - 320 * 320 pixels, nowhere near what people expect from even a basic phone camera these days - but to get high resolution images required a very expensive upgrade. What was harder work than we expected was integrating the camera into the NVIDIA Jetson Nano systems that we installed on-premise. But it was proper IR, it did not need an illumination source and it provided additional information to the visual light cameras that the machine learning algo could use to assist with the counting process.

 

[Drakkith]

I was surprised by the low resolution - 320 * 320 pixels, nowhere near what people expect from even a basic phone camera these days - but to get high resolution images required a very expensive upgrade.

I think it has something to do with the noise of these uncooled microbolometer sensors (I assume that's what type it was). According to wiki, their design makes them far more susceptible to noise at small pixel sizes compared to cooled thermal cameras. And large pixels usually means less resolution.

 

[Melbourne Guy]

I think it has something to do with the noise of these uncooled microbolometer sensors

I'd have to pull the spec sheet to recall the sensor type, but yes, @Drakkith, noise was a real problem. In our testing, the ML could tell an object was a person, but that was about it, such was the limited discrimination of the images. (Training the machine was also tedious because it is hard to find pre-existing IR models of the kind we needed, but now I'm rowing hard in the opposite direction of the OP, so that's all I'll say about that!).

It did make me wonder about the practicality of low cost IR goggles, though. It's a blurry view of the world, so I guess it would be okay in very dark environments, but you'd still be likely to trip over things because small objects would be hard to make out as you moved.

 

[Rive]

low cost IR goggles

On the market, the business name and the principle it's working with is not necessarily fits each other. A 'low cost IR goggle' is usually some night vision stuff, based on an image amplifier: or some IR sensitive camera with illumination. No real IR (in sense of thermal) there.

Ps.: things are worse than I've thought. For example on Amazon, practically everything I've just found below $400 is a cheap camera + IR torchlight. And a screen. With deceptive description and questionable quality...
The first real thermal thing is at $550, with resolution of 160*120...
Things with an actual image intensifier seems to start above $1000, but often has no price displayed and comes with US export restriction...