Public Lab Research note

Single camera NDVI solution

by theowallis | August 08, 2014 16:50 08 Aug 16:50 | #11022 | #11022

What I want to do

I have a Mobius Actioncam from which I have removed the ir blocking filter. I'm aware that a single camera with a blue filter is an option for pseudo NDVI images. The option for true NDVI images with two consumer cameras as far as I'm aware is to fit one camera with an infrared filter and leave the other camera as a visible spectrum camera so you get the RGB & NIR bands.

I'd like to see if there is a way to isolate, extract and save as three individual bands the NIR, red and green bands from a single camera with only the ir blocking filter removed as it (within the tolerances of the sensor) captures the full light spectrum.

My attempt and results

My attempt so far is only to remove the ir blocking filter from my camera so that it now takes full spectrum images.

Questions and next steps

How do I go about extracting and saving the ~700nm to 800nm NIR band with software? I have looked into this but am yet to find a solution.

Why I'm interested

The intended use of this camera would be for analysing vegetation stress.


That's a tough one. With the IR blocking filter removed, all three color channels (all pixels, regardless of which color array filter they are under) will capture some near infrared light. So the camera records three channels, each with mostly information about red, green, or blue light, but also with some unknown amount of NIR (and some UV as well). Trying to figure out how much of that brightness in each channel is from NIR light might be futile. Here are a couple examples of what the three color channels see when the IR block filter is removed (but each camera is different):


That's why single camera NDVI systems use a filter. A red filter that lets no blue light through allows only NIR light (and some UV) to reach the blue channel (the pixels under blue CFA filters). The red and green channels are a mix of visible and NIR light, but it still sort of works. To get a mostly pure NIR channel and a mostly pure red channel, two cameras are needed.


Reply to this comment...

Thanks for your comment, this is what I was afraid of; there is no relatively simple way to isolate specific wavelengths and furthermore if this was done those wavelengths that were to be used for red and green would contain infrared information thus providing false (or at least less accurate) NDVI readings. Hmmm, looks like it's the two camera system which would produce more true NDVI readings than one camera with a blue filter.

Thanks again!

Reply to this comment...

What about if you added a Schott BG3 filter behind or in front of the lens, then you would have an infrablue response curve (mostly IR in the red channel, mostly visible in green and visible+IR in blue channel).

Reply to this comment...

Login to comment.