How to detect a specific color in photo?

Hello. Recently, I faced some problems in using the image analysis. In attached photo, I needed to calculate the percent of shell(or root, sediment etc.) in total area. I already tried to detect the shell and calculate the percent of shell in total circle area.

Then, I plotted the red marker at each point, and could calculate the percent of shell. The following is code to detect the shell.

rgbImage = imread('GEMS_Vegetation.png');
imshow(rgbImage);

[x,y] = ginput(1);
    row = round(y);
    column = round(x);

r = rgbImage(:,:,1); r_dou = im2double(r)*255; Red   = r_dou(row, column);
g = rgbImage(:,:,2); g_dou = im2double(g)*255; Green = g_dou(row, column);
b = rgbImage(:,:,3); b_dou = im2double(b)*255; Blue  = b_dou(row, column);

Error_R = 0.15;
    [R_r R_c] = find(r_dou <= Red+(Error_R*Red) & r_dou >= Red-(Error_R*Red)); R = [R_r R_c];
    [G_r G_c] = find(g_dou <= Green+(Error_R*Green) & g_dou >= Green-(Error_R*Green)); G = [G_r G_c];
    [B_r B_c] = find(b_dou <= Blue+(Error_R*Blue) & b_dou >= Blue-(Error_R*Blue)); B = [B_r B_c];

[RG,ia,ib] = intersect(R,G,'rows'); [RGB,ic,id] = intersect(RG,B,'rows');

figure()
    imshow(rgbImage); hold on;
    plot(RGB(:,2), RGB(:,1), 'ro', 'markersize', 3);

T_color = find(r_dou ~= 0);
Obj_pixels = (length(RGB)/length(T_color))*100;

However, in detecting the roots(greenish color) or stem, we couldn't detect the proper values. I thought that there is a water, inducing a reflection and shadow.

In this point, how can I properly detect the wanted colors..??

Answer 1

[Kushagr Gupta]

I would recommend the following link to do image segmentation in color space:

https://www.mathworks.com/help/images/examples/color-based-segmentation-using-k-means-clustering.html

Moreover, there is a imageSegmenter app in MATLAB that can be used for trying out different segmentation techniques in a quick way to see which performs the best. It can be accessed by executing the following command:

>>imageSegmenter

Hope this helps!