Try to extract the RGB channels of the image (you can do it using R and EBImage library and the function channel); then recompute the channels as follows:

r= R/(R+G+B), g= G/(R+G+B), b=B/(R+G+B); this pre-processing is useful to remove scene illumination influences;

the final step is to calculate features representing greenness as follows:

Grenness=2G-R-B

or

Grenness=2G+B-2R

Both methods are able to isolate Green tissues from residue/soil background.

The obvious answer would be:

- Separate the image in RGB channels.

- Green is where the G pixels are high and the R and B pixels are low.

- If forests are green the whole year (tropics), you could eliminate them because they stay more or less constant in the time-series.

You may use a professional (still free !) software package called CAN-EYE developed by INRA Avignon (France): http://www6.paca.inra.fr/can-eye. You can retrieve the green Leaf Area Index, and not the plant chlorophyll content.

You may not really be able to get the chlorophyll content from camera photograph, what is teh intention to work on Phenology?

Thank you @Lambert and @Ruxandra.

Yes this work relates to phenology and Thank u @Saif

@Lambert Zijp:...Separate the image in RGB channels.

In Matlab, try

h = imread('onion.png'); %built-in image for experiments

g = h(:,:,2); % green channel values

subplot(2,2,1); imshow(h); title('onion.png');

subplot(2,2,2); imshow(g); title('green parts of onion.png');

subplot(2,2,3); plot(h); title('average colour values plot');

subplot(2,2,4); plot(g); title('green intensities plot);

@James Peters : just evaluating the G channel is not enough. White areas (e.g. sandy beaches) also have high values in G, so I would add as criterion that besides G being high, the R and B should be low enough.

James will this highlight chlorophyll.

Lambert,

Good post. I agree with your suggestion.

Hi Jayan,

I developed a method which allows you to estimate chlorophyll at leaf level using a commercial RGB camera. The method is explained in a poster which you can download at the URL here below:

https://www.researchgate.net/profile/Frank_Veroustraete/publications/?pubType=inProceedings&ev=prf_pubs_inp

Mind that this method has been developed for leaf level measurements, not for canopies. However you can perform up-scaling of chlorophyll to (rice) canopy level by measuring a sufficiently large set of rice plant leaves and subsequently apply a canopy model using an erectophile leaf distribution (for rice plants). Give it a try, I think for homogeneous canopies like with rice plants it should work out fine.

Cheers,

Frank

PS.: You need a leaf and canopy model (is explained in the poster) to make your method specific for rice plant canopies. If you want I can provide the models which I used for this purpose.to determine the chlorophyll content of Tilia leaves and trees.

I would also suggest some indices for green extraction that you can find in these articles:

Meyer, G.E. & Neto, J.C. 2008, "Verification of color vegetation indices for automated crop imaging applications", Computers and Electronics in Agriculture, vol. 63, no. 2, pp. 282-293.

Casadesus et al. 2007 Using vegetation indices derived from conventional digital camera as selection criteria for wheat breeding in water-limited environments. Annals of Applied Biology 150, pp. 227–236

@Frank thank u very much and it is nice to hear ur work. I will write to u.

@Daniele I will go with that paper and thank u very much for your kind support

Try to extract the RGB channels of the image (you can do it using R and EBImage library and the function channel); then recompute the channels as follows:

r= R/(R+G+B), g= G/(R+G+B), b=B/(R+G+B); this pre-processing is useful to remove scene illumination influences;

the final step is to calculate features representing greenness as follows:

Grenness=2G-R-B

or

Grenness=2G+B-2R

Both methods are able to isolate Green tissues from residue/soil background.

@Francesco Chianucci: following up on your suggestions, I found better results using the hue saturation value (hsv) colour space instead of the rgb colour space. Here are the steps:

1. convert an rgb image to an hsv image.

2. obtain almost pure hsv values for each colour using your first suggestion.

Here are the results I obtained using Matlab:

% hsv to pure hsv segments

clc, clear all, close all, whitebg('w')

g1 = imread('peppers.png');

%

h1 = rgb2hsv(g1);

h1r = h1(:,:,1)./(h1(:,:,1)+h1(:,:,2)+h1(:,:,3));

h1g = h1(:,:,2)./(h1(:,:,1)+h1(:,:,2)+h1(:,:,3));

h1b = h1(:,:,3)./(h1(:,:,1)+h1(:,:,2)+h1(:,:,3));

%

subplot(2,3,1); imagesc(h1(:,:,1));title('red segment');

subplot(2,3,2); imagesc(h1r);title('hsv red channel');

subplot(2,3,3); imagesc(h1(:,:,2));title('green segment');

subplot(2,3,4); imagesc(h1g);title('hsv green channel');

subplot(2,3,5); imagesc(h1(:,:,3));title('blue segment');

subplot(2,3,6); imagesc(h1b);title('hsv blue channel');

The attached image shows the results.

@Francesco Chianucci: following your first suggestion using only the rgb colour space, I used the following Matlab script to obtain pure channel images:

% rgb to pure colour segments

clc, clear all, close all, whitebg('w')

g1 = imread('peppers.png');

%

g1r = g1(:,:,1)./(g1(:,:,1)+g1(:,:,2)+g1(:,:,3));

g1g = g1(:,:,2)./(g1(:,:,1)+g1(:,:,2)+g1(:,:,3));

g1b = g1(:,:,3)./(g1(:,:,1)+g1(:,:,2)+g1(:,:,3));

%

subplot(2,3,1); imagesc(g1(:,:,1));title('red segment');

subplot(2,3,2); imagesc(g1r);title('rgb red channel');

subplot(2,3,3); imagesc(g1(:,:,2));title('green segment');

subplot(2,3,4); imagesc(g1g);title('rgb green channel');

subplot(2,3,5); imagesc(g1(:,:,3));title('blue segment');

subplot(2,3,6); imagesc(g1b);title('rgb blue channel');

The results produced by this script are shown in the attached image. The deep red segments, for each channel, exhibit the pure colour channel brightness values.

@Francesco Chianucci: following your second suggestion using only the rgb colour space, I used the following Matlab script to obtain less satisfactory results:

% rgb to pure colour segments (method 2)

clc, clear all, close all, whitebg('w')

g1 = imread('peppers.png');

%

g1r = g1(:,:,1)./(g1(:,:,1)+g1(:,:,2)+g1(:,:,3));

g1g = g1(:,:,2)./(g1(:,:,1)+g1(:,:,2)+g1(:,:,3));

g1b = g1(:,:,3)./(g1(:,:,1)+g1(:,:,2)+g1(:,:,3));

g = 2.*g1r-g1g-g1b; r = 2.*g1g-g1r-g1b; b = 2.*g1b-g1g-g1r;

%

subplot(2,3,1); imagesc(g1(:,:,1));title('red segment');

subplot(2,3,2); imagesc(g);title('rgb red channel');

subplot(2,3,3); imagesc(g1(:,:,2));title('green segment');

subplot(2,3,4); imagesc(r);title('rgb green channel');

subplot(2,3,5); imagesc(g1(:,:,3));title('blue segment');

subplot(2,3,6); imagesc(b);title('rgb blue channel');

You can see from the attached image that the red and blue channel displays contain impurities. However, the green channel results are remarkable, since the pure green channel values are illuminated with only minor red impurities (the blue channel display can be ignored).

Thanks James and Francesco for excellent contributions.

@James Thank you for kind advice and ideas to do this using MATLAB. It will helpful to everybody. @Francesco Thank u for your update regarding R.

didn't see the original image