It is impossible for j to go larger than numel(D) unless you modify j or D within the loop (which you are doing). I believe what's happening is that when you run the for loop, Matlab doesn't recompute numel(D) each iteration; it's computed only once initially. Your code then erases elements of D, making it have fewer elements. So initially, D may have 10 elements. As you iterate, say you delete two of those elements. When the loop gets to j = 9, there are only 8 elements so you get the error.
A better way to do this is to keep track of the indices you want to remove, then remove them all at once. This also has the advantage of not skipping indices (for example, if j = 5 and you remove D(5), then D(6) becomes D(5), but on your next iteration j will be 6, skipping over the new D(5).
for i = 1:numel(B)
remove_these = [];
for j = 1:numel(D)
dist = sqrt((B(i).x - D(j).x)^2 + (B(i).y - D(j).y)^2);
if dist < B(i).size/2 & B(i).size/2 > D(j).size
B(i).size = B(i).size + D(j).size;
remove_these = [remove_these j];
end
D(remove_these) = [];
end
end
Finally, I don't know what else your structure stores, but it may be more convenient just to use matrices if you're just storing x,y coordinates. Makes indexing easier.
