So NDVI or not?

Posted by Quentin Mayberry on September 5, 2015 at 1:46am

As we know there has been a lot of discussion on this subject and no clear winner from what I can see.

We are looking at the micasense red edge sensor, but it is a huge capital outlay for several units, so naturally apprehensive.

We have been playing around with a sample field for a farmer just using a straight forward RGB camera but just upped the temperature of the image...

Sample one taken 14/3/2015

3691241389?profile=original

Then another shot of the same field on 12/4/2015

3691241522?profile=original

Then another on 16/7/2015

3691241427?profile=original

Interesting results, obviously the multicopter camera angle changed, but the farmer was able to determine the areas that were poor and was able to get a rough idea of how much crop he could obtain from this field as the dimensions are known.

So is there software out there that will calculate the areas of the image which we can calibrate to give us a yield from the imagery without too much messing about?

And is NDVI really necessary when we get very similar results from straightforward RGB imagery?

You need to be a member of diydrones to add comments!

Join diydrones

Email me when people reply –
Follow

Replies

  • RG September 5, 2015 at 9:33am

    Hi Quentin, idea of using NDVI is to detect changes very early on which might not be captured in RGB. Its not MUST have and depends on ROI and how sophisticated you would like to get. We have a drone image processing service where we can provide you with approx. yield from your images along with creating ortho (stitching of photos), DEM, NDVI etc... if you would like email me at info@harvesting.co 

    • Quentin Mayberry > RG September 5, 2015 at 1:27pm

      Thanks for the information RG, I'll be in touch.

  • Antonio September 5, 2015 at 8:10am

    OpenCV might help you to do the work through counting pixels of a threshold image, as long as you can establish a scale.

    I don't know of any ready solution.

    • Quentin Mayberry > Antonio September 5, 2015 at 1:26pm

      Thanks Antonio, I'll have a look at that.

  • Marc Cretten September 5, 2015 at 3:02am

    Interesting, how did you up the temps of the RGB? Why not convert a cheap camera to NIR? We in the same boat as you with the micasense...its pricey. (Following this thread)

    • Quentin Mayberry > Marc Cretten September 5, 2015 at 4:23am

      Simply select maximum temperature of the image using lightroom Marc.

      We haven't found any advantage of the NIR, as we have tried that.

  • Antonio September 5, 2015 at 2:54am

    kI Quentin

    Of course that there are lots of edge tech that apply when it comes to knowledge extraction but you have to attend to ROI's and alike indicators, and for all that really matters try to stick to KISS principle..

    So, I would dare to say that if the main interest is to have a snapshot from a single monoculture field as to know how crops are evolving in terms of growth and irrigation a straightforward RGB imagery will do it.

    I guess that when it comes to pragma it is like the song: "The only thing that's worst then one is none"

    • Quentin Mayberry > Antonio September 5, 2015 at 4:25am

      Yes we are just interested in grass yields for a current project, not multi crops, interesting response :) 

This reply was deleted.