Infragram Near-Infrared Wavelengths

[zbarr...@gmail.com]

Hello,

I understand that the Infragram is a modified camera with the infrared filter removed, and the resulting near-infrared channel replaces the red channel during processing and display.

Can somebody please tell me what range of near-infrared wavelengths are "unlocked" with the use of the Infragram webcam? Which red wavelengths are omitted?

The site mentions that digital cameras are able to see into the near-infrared, but those wavelengths are filtered out so that only the visible spectrum remains. If somebody could please tell me this information or point me to a relevant data sheet, that would be great.

Thank you very much,
An Interested Infragram User

[Chris Fastie]

Dear Interested,

Single-camera infrared systems like the Infragram have the factory infrared blocking filter removed and are then capable of capturing near infrared light up to about 1000 nm. But they are not very sensitive to NIR longer than about 850 nm. The following is a very generalized diagram. Values will be different for each camera sensor, and the cameras sold by Public Lab are CMOS not CCD.

When a blue filter replaces the IR block filter as it does in some Public Lab Infragram cameras, red light and most green is blocked, so mostly NIR is captured in the red and green channels:

When a red filter replaces the IR block filter as it does in some Public Lab Infragram cameras, blue light and most green is blocked, so mostly NIR is captured in the blue and green channels:

Still interested?

Chris

[zbarr...@gmail.com]

Chris,

Wow you are quick. Thank you for the speedy reply.

I have the Infragram webcam.
http://store.publiclab.org/collections/diy-infrared-photography/products/infragram-webcam

After reading your insight, then it looks like the webcam has a blue filter (with most red and green being blocked, and NIR being displayed in those channels).

"allows you to take an infrared photo in the "red" channel of your camera, and a visible image in the "blue" channel."

I am interpreting the first graph as the blue line representing the relative response when a blue filter is being used. In the third graph (blue filters), it looks as though some filters don't extend past the visible.

I think I'm onto it now, Chris. Your help is greatly appreciated.

Please let me know if I'm interpreting your insight correctly and if indeed the Infragram webcam uses a blue filter.

I could use some help understanding the post-processing methods done to achieve optimal NDVI contrast. Do you think you'd be able to help me with this?

Thanks for your time,
Zach

[Chris Fastie]

Yes, I think the Pubic Lab webcam Infragram camera has a blue filter (Rosco 2007?) in it. 

The blue line in the first graph (below) is sort of "the relative response when a blue filter is being used." 

The blue line is actually the response of the blue channel when the IR block filter has been removed. The blue channel captures lots of blue and NIR light. If you add a blue (e.g., Rosco 2007) filter which blocks red and some green light, the blue response of the camera is not changed very much because the blue channel is not capturing red and green in the first place. Remember that this is a very generalized diagram for a CCD (not CMOS) sensor.

There is some post processing that can improve NDVI images. However, I have not seen any examples of meaningful NDVI images made from photos taken by a blue filtered CMOS camera that has not been custom white balanced. It might be possible to find a program that allows controlling the white balance (see this thread). Without white balance control, it might be difficult to get useful NDVI results from that camera. Even with custom white balance, good results could elude you. Sometimes blue filters just don't work in inexpensive CMOS cameras. I don't know if anyone has ever tried a red filter in that camera.

Chris

[zbarr...@gmail.com]

Chris,

Your insight is helping me a lot here.

I had seen that thread and have read over several relevant threads to help with my understanding. I have also read over some of your notes ("Orange is the New Red," "NIR and White Balance," among others). I really enjoyed that white balance note that you wrote some months back.

I now want to understand the post-processing of NDVI. I still need to familiarize myself with the Python codes already available; I do understand that the photo taken by the camera is not the goal, rather the data of the arithmetic index for each pixel.

A common formula I am seeing is (R-B)/(R+B). Do the "R" and "B" simply represent the brightness value of the respective field in each pixel (i.e. brightness value of red and blue in however many bits)?

An Infragram webcam with a blue Rosco 2007 filter will block red light and most green, as you said, and the field values in those channels will represent mostly NIR data. (The blue is the best data we have to represent what ideally would be the spectral reflectance of all visible light, but again, this is only a one-filter camera.)

As far as white balance goes, I saw that you placed a piece of "blue paper under blue sky in the shade." During your white balance customization, were you trying to make the piece of paper appear as white? Something else?

It seems rather intuitive whatever you did do. Maybe we can get reasonable NDVI results from the Infragram webcam if I customize properly.


Does Public Lab have any data sheets available for the equipment on the Infragram webcam? I see that a company was assigned to attach webcams (no IR filter, added Rosco) to the circuit board. Do you know what the name of the CMOS sensor webcam is? Maybe I can find some specs on it.

Thanks,
Zach

[Chris Fastie]

 

A common formula I am seeing is (R-B)/(R+B). Do the "R" and "B" simply represent the brightness value of the respective field in each pixel (i.e. brightness value of red and blue in however many bits)?

That formula will approximate NDVI if NIR is being captured in the R channel and visible light is being captured in the B channel. That is more or less what a blue filtered camera does. In an 8 bit image (like most jpegs from consumer cameras) each pixel has a number from 0 to 255 for each color channel (RGB) representing the brightness of that color.  That formula will use those numbers to produce an index (unitless value) which highlights the difference between the brightness of NIR and the brightness of blue. NDVI is traditionally computed with physical measures of the radiance or reflectance of foliage, and the brightness recorded by cameras is not related to those measures in very predictable ways.

 (The blue is the best data we have to represent what ideally would be the spectral reflectance of all visible light, but again, this is only a one-filter camera.)

NDVI is traditionally computed using the energy of red wavelengths to represent visible light. Plants absorb all wavelengths of visible light for photosynthesis, so any visible wavelength will tell us something about how much photosynthesis is happening. But red light has some advantages and legacy NDVI typically uses red.

 

As far as white balance goes, I saw that you placed a piece of "blue paper under blue sky in the shade." During your white balance customization, were you trying to make the piece of paper appear as white? Something else?

Doing a custom white balance while flooding the sensor with blue light fools the camera into making severe adjustments to correct for the detected color imbalance. The camera then blindly dials down the blue and dials up the red for every pixel of every photo. This exaggerates the brightness of the red channel for every photo, and we use the red channel to represent NIR. If you get the white balance procedure right, the difference between the blue value and the red (NIR) value for plant pixels will compute to a reasonable facsimile of NDVI.

 

 Maybe we can get reasonable NDVI results from the Infragram webcam if I customize properly.

The inexpensive sensors in some little webcams have some strange responses to blue filters.  If you can control the white balance setting, you might get good results.

  

Does Public Lab have any data sheets available for the equipment on the Infragram webcam? I see that a company was assigned to attach webcams (no IR filter, added Rosco) to the circuit board. Do you know what the name of the CMOS sensor webcam is? Maybe I can find some specs on it.

I don't have those answers. 

 

[zbarr...@gmail.com]

Chris,

Thanks for all the very good information.

Based on our discussion, what is your recommendation for a camera and filter combination that is lightweight (for aerial use), inexpensive, and capable of reasonable NDVI results?

Thank you,
Zach

[Jeffrey Warren]

There should be datasheets at the bottom of the infragram webcam wiki page!

--
Post to this group at plots-i...@googlegroups.com

Public Lab mailing lists (http://publiclab.org/lists) are great for discussion, but to get attribution, open source your work, and make it easy for others to find and cite your contributions, please publish your work at http://publiclab.org
---
You received this message because you are subscribed to the Google Groups "plots-infrared" group.
To unsubscribe from this group and stop receiving emails from it, send an email to plots-infrare...@googlegroups.com.

[Chris Fastie]

Lightweight and inexpensive are rather subjective terms. A Mobius Actioncam is so light it can fly on anything anytime. The default lens is really wide angle, but that might be what you need. If you get the white balance and exposure right, it can make fine NDVI images.

A Canon PowerShot can cost much less than a Mobius and has far superior lens and sensor quality so the photos are much higher resolution and crisper. The CCD Powershots seem to make NDVI images with better dynamic range than the Mobius. PowerShots weigh as little as 135 g, but the Mobius weighs only 38 g.

Any of the PowerShots that are easy to convert to IR might be a good choice. Either a red or orange filter can work well. Sometimes you can buy a refurbished Canon PowerShot S100 (really nice) which is easy to convert for less money than the Infragram Point and Shoot Mobius.

[zbarr...@gmail.com]

Thanks Jeff.

Zach

[zbarr...@gmail.com]

Thanks Chris.

Zach