I had time for a quick research project about using CIR(Near infrared) cameras, RGB cameras, and what those sensors can really provide for mapping vegetation. This article's purpose is to show the value of using Color Infrared (Near Infrared) sensors, and some of the misconceptions about just needing a visible light (RGB) camera.
Thanks for some great quotes from:
- Dr. Kevin Price (Agpixel, Iowa State University)
- Dr. Deon VanDerMerwe (Kansas State University)
- Gabriel Torres at Micasense
- Beau Dealy at APIS remote sensing
Here's a link to the article:
I really could use some help getting my NDVI images to look like correct NDVI images. I'd prefer the Green=High NDVI. I modified my Canon Powershot sx280—Took out the hot mirror and put in the Midopt DB660/850 filter.
I guess I just don’t know how to work the Fiji/ImageJ software well enough. Below is what I’m getting. My first picture (Auto Settings)—and where I’m getting hung up.
What are the correct ImageJ settings for the DB660/850 filter? How can I get the image to show high NDVI in GREEN??
That's a good point to distinguish too Robert, difference between a good large sensor and an actioncam on a Phantom.
Keep in mind that many plant stressors (e.g. nutrient deficiency) will have the largest and earliest impact by decreasing the concentration of absorbing pigments that will be detected in the visible range as increased reflectance, especially the red. In our Arctic vegetation mapping work, we’ve had success getting decent RGB dynamic range in UAV imagery using a Sony a6000 that has a large sensor…but as you suggested Kyle I’m not sure that we could get a similarly useful visible range with a GoPro or Phantom.
If there is some function that accepts only the image in the upper left and converts it to the image in the upper right, then that function is written wrong because I can clearly see that the best, greenest areas are marked as the areas of the poorest health. Like Ned says, if someone can perceive the differences in an RGB image, they can create an index that shows those differences. I can see that the areas marked as good in the lower right image are clearly greener in the RGB image than the areas marked as bad. Therefore, I can create an index that thoroughly outperforms whatever index you're using to create your false NDVIs.
So, when you say that false NDVIs aren't as good as CIR-based NDVIs, I think to myself that maybe *these* false NDVIs aren't as good, since they're clearly poor (since they show clearly green crops as being in poor heath), but that is not a good argument that a good false NDVI is not as good as a false one from RGB data.
I think it's important to explain why the false NDVI in this post shows what look like very healthy areas to any untrained observer as being in bad health compared to the surrounding area.
Just as an FYI, I didn't create the False-NDVI index. I agree that this index should be showing up healthier. But a company that processes A LOT of agricultural maps came up with this index, and I uploaded some imagery to give it a try. They need to work on their indexes, but ultimately, no matter how good the index from RGB is, you can't make up for having NIR simply from the reflectance standpoint
That can happen when the reflectance is so low. From previous flights with multiple RGB cameras, some shades of green don't reflect as much as other shades of green. So it may be more green, but the camera is so sensitive that it doesn't register as green as it truly is. Different varieties of wheat, corn, soy, barley, etc. are different shades of green, so that can really mess with the scale when it is so small
The original post is missing some important details. It would be helpful to know if the images were calibrated or ocntroled in any way before creating the NDVI and false-NDVI images. For any index to be valid some calibration or imaging protocol must be in place or the index can result in nonsense. False-NDVI is a new term for me. I've been working with NDVI for about 35 years and I've seen more variations and misinterpretations of NDVI in the last two or three years then the previous 30+.
In general if you can see a difference in an RGB image then you can be sure that you can create some sort of index that represents that difference. You can produce reasonable vegetation indices from RGB but they are not as robust or sensitive as metrics such as NDVI. For what it's worth, it seems as it most people playing with converted cameras (myself included) create an NDVI image and then compare it to an RGB image to see if the results are any good.
If the false NDVI in this post (top right image) was generated solely from the RGB image in this post (top left image), why does it show poorer crop health (red) for areas that are clearly greener in the RGB image?
This article was coming from the perspective of a grower (farmer). They may fly over the crops more than once, but they only look at one dataset at a time. So in agriculture, right now the "False-NDVI" is mainly used from one flight of an RGB (usually from a phantom camera, or something similar).
The reason why this article was written is because of misinformation. This index was created to look like an NDVI, but you don't have to pay extra to buy modified cameras. Farmers think that this False-NDVI is just as good, but really they are getting back inaccurate results
Is a false NDVI created using a single RGB image, or an RGB image at test time plus another RGB image taken at some earlier reference time?