Agisoft Metashape

Agisoft Metashape => General => Topic started by: IJohnson on May 20, 2014, 10:33:29 PM

Title: Color correction and NDVI
Post by: IJohnson on May 20, 2014, 10:33:29 PM
Hi,
Is anyone using photoscan to merge Near infrared images?  I'm looking to merge raw NIR images together before NDVI analysis, and part of my regular (visual spectrum only) workflow is to use color correction to balance out any exposure issues that occurred in flight. 

My concern is that Photoscan's Color Correction feature may alter individual RGB channels separately rather than uniformly, which could affect NDVI readings.  A uniform or luminosity-only correction wouldn't be an issue though, given that NDVI is a normalized index. 

So my main question is this- does Photoscan's color correction feature affect individual color channels, thereby breaking NDVI workflows?
Title: Re: Color correction and NDVI
Post by: nprokofyev on May 23, 2014, 03:24:27 PM
Aerie,
did you know that most of CMOS sensors are sensitive to NIR on either R, G1, G2 and B channels? If you have photos made thru the IR filter, just desaturate them. >95% of resulting "luminosity" is pure NIR.
As for RED, I think it is also not a problem to desaturate photos except one case. It is when smb using CMOS R channel instead of photos with red filter.
IMHO
Title: Re: Color correction and NDVI
Post by: IJohnson on May 23, 2014, 04:49:14 PM
Aerie,
did you know that most of CMOS sensors are sensitive to NIR on either R, G1, G2 and B channels? If you have photos made thru the IR filter, just desaturate them. >95% of resulting "luminosity" is pure NIR.
As for RED, I think it is also not a problem to desaturate photos except one case. It is when smb using CMOS R channel instead of photos with red filter.
IMHO
I did not know this.  Do you have any examples or research you can pass along to help explain your findings?
Title: Re: Color correction and NDVI
Post by: nprokofyev on May 29, 2014, 04:39:48 PM
  Do you have any examples or research you can pass along to help explain your findings?
Well, I am developing custom photosetup for NDVI, and have some sample photos in NIR with 720, 760 and 950-nm cutoff filters.
I can make a RAW photo of cigarette lighter in dark room, and you will see, all RAW channels will contain data.
Title: Re: Color correction and NDVI
Post by: nprokofyev on May 29, 2014, 06:26:44 PM
This is the IR LED in dark room. Photo with 760 nm filter.
sorry, had to cut, if you need a full-size raw, ask me in PM.
Title: Re: Color correction and NDVI
Post by: IJohnson on May 30, 2014, 05:24:11 PM
Very interesting research, thanks for sharing.
Title: Re: Color correction and NDVI
Post by: nprokofyev on June 19, 2014, 12:32:29 PM
Aerie, could you explain in a few words how to calculate reflectivity? For example, if you have NIR photo, the measured parameter is pixel brightness. How to calculate reflectivity then?
Title: Re: Color correction and NDVI
Post by: frank.stremke on June 23, 2014, 11:39:56 PM
This sounds very interesting
Can you give some more derails on how to do it iwould like to give it a try myself
Thanks
Frank
Title: Re: Color correction and NDVI
Post by: IJohnson on June 23, 2014, 11:44:07 PM
nprokofyev - We calculate plant stress by using the Normalized Differential Vegetative Index, which is a simple index that compares the brightness of a visible band (red for satellites, usually blue for UAVs) to the brightness of the NIR band.  See here: http://en.wikipedia.org/wiki/Normalized_Difference_Vegetation_Index

But keep an eye on DIYDrones.com, there are some interesting developments going on over there in attempts to develop a more reliable index.  NDVI is susceptible to false readings (especially from shadows), which can be more of an issue for UAV operators because of the increased resolution.
Title: Re: Color correction and NDVI
Post by: nprokofyev on June 25, 2014, 10:08:43 AM
Well, Aerie, NDVI calculation using brightness indeed is susceptible to false readings. And the open problem that is in front of me is to develop measurement method that operates reflectivities (as it is defined for NDVI in wikipedia) instead of brightness.
Reflectivity is a property of material and is independent to sun brightness, shadows, etc.

Anyway, you cannot calculate true NDVI simply comparing brightness of photos made in RED and NIR range.