How to detect and count the tree crown with a very high density automatically?

I'm trying to detect and count the palm trees at low density and high density. I used CircularHough function to detect circular radius and circular centroid of palm trees after pre-processed image. I got a fairly good result but there are some errors in the detection and some undetected trees which high density. My result showed in the link below.

I used this code to detect the tree crown.

tic;
[accum, circen, cirrad] = ...
    CircularHough_Grd(a_dilate1, [5 22],...
    8, 13, 1);
toc
if any(cirrad <= 0)
    inds = find(cirrad>0);
    cirrad = cirrad(inds);
    circen = circen(inds,:);
end
togglefig Results
imshow(a_dilate1);
hold on;
plot(circen(:,1), circen(:,2), 'rh');
for ii = 1 : size(circen, 1)
   rectangle('Position',[circen(ii,1) - cirrad(ii), circen(ii,2) - cirrad(ii), 2*cirrad(ii), 2*cirrad(ii)],...
        'Curvature', [1,1], 'edgecolor', 'b', 'linewidth', 1.5);
end

My question is why this code can't detect the tree crown with higher density? Anyone can help me, please? And how to detect the tree crown with a very high density as in the picture below?

Answer 1

[Image Analyst]

If the image is pretty much solid green, how do you expect it to find the trees? Maybe you can just go with area fraction rather than the count.

[Azmel]

Is it possible to detect the trees with cropped one tree as a sample image or by shape of tree crown?

[Image Analyst]

Tree counting is not my field. But it is for people who have published papers on it that are listed in section 22.5.8 here: http://www.visionbib.com/bibliography/contentscartog.html#Cartography,%20Aerial%20Images,%20Remote%20Sensing,%20Buildings,%20Roads,%20Terrain,%20ATR