Question: NDVI validation

aldrin is asking a question about ndvi
Follow this topic

by aldrin | October 10, 2016 01:33 | #13544


How can i validate that the data I obtained using raspi noir camera are really NIR and NDVI?

I am still new in this topic, and we were tasked to create a low cost ndvi camera, I understand that using pi noir camera with blue filter, we could get false color chanel that gives NIR(instead of rgb, it would be NIR,NIR,Blue+NIR). My problem is that how can i show or validate that what my camera really produce is NIR or NDVI? I found one multispectral camera in one of the universities near our school, can i use that on validation? And also my professor wants me to show the standards that can prove that NDVI is really acceptable. Please Help me,



4 Comments

Depending on the blue filter used, photos from a Pi NoIR camera would capture (instead of R,G,B) NIR+R, NIR+G+B, B+NIR+G. If you use a good red filter that really blocks all blue and green, you could capture R+NIR, NIR+R, NIR. Some little CMOS sensors give poor NDVI results with a blue filter compared to a red filter, but I don't know about the Pi NoIR camera.

There are lots of DIY tests to learn what is being captured in each channel:

  • Take a photo of an NIR LED (or remote control) that cannot be seen with the naked eye. Which channels record the LED? In which channel is it brightest?
  • Take a photo of the spectrum of a compact fluorescent lamp created with a slit and piece of DVD. Which channels record the blue, green, and red peaks?
  • Take a photo of healthy green foliage outside. In which channels does the foliage look brightest? That's the channel with the most NIR.

Do these tests with a normal Pi camera and Pi NoIR camera both with and without a blue filter and then red filter. You will know more than most people about what is going on.

Your professor should not be satisfied that "NDVI is really acceptable" until you do what @claytonb did: https://publiclab.org/notes/Claytonb/08-13-2016/plant-health-ndvi-white-balance. All you need is a Canon S100 and some Boom Chicka Pop (and an expensive filter).

Check out these notes (and comments) about Pi cameras and NDVI:
https://publiclab.org/questions/Kiranay/10-15-2015/question-determining-plant-health
https://publiclab.org/notes/khufkens/11-02-2015/ov5647-raspberry-pi-camera-spectral-response-quantum-efficiency
https://publiclab.org/notes/LaPa/03-31-2016/raspberry-noir-cam-sensors-to-detect-water-stress-of-the-plants-during-their-growing

Chris

Is this a question? Click here to post it to the Questions page.

Thank you, so much. I'm planning to use DB660/850 which is a Dual Bandpass filter for NIR and VIS. I understand that if I use it, I can Get Red+NIR, NIR and NIR from the original RGB channel. But, is there way to separate the NIR from red red channel to get a pure red?

Is this a question? Click here to post it to the Questions page.


Even with that nice filter it won't be possible to get a pure red channel. The red channel is sensitive to NIR light, and NIR is transmitted by the filter, so there will be NIR mixed with the red. There is a discussion of this here.

Some calibration procedures assume that a constant percentage of the red channel is NIR and subtract that proportion from the red channel for each pixel. Determining what percentage to use is a challenge unless you know the spectral sensitivity of the sensor or have a valid NDVI image of the scene or some other knowledge to use for calibration. A bigger problem might be that the proportion of NIR in the red channel is not likely to be the same for every pixel in the scene. Some pixels will have received no red at all, some pixels will have received no NIR, and other pixels could have a wide range of mixtures. But for green foliage in daylight, it might be a good assumption that there will be a more or less constant proportion of red and NIR.

Chris


Reply to this comment...


Hi - would you mind making the title "How can i validate that the data I obtained using raspi noir camera are really NIR and NDVI?" so it's easier to understand the question in list view? Thanks!

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Log in to comment