How to place text label near line, but avoiding other lines and labels?

Question

KAE 2021-6-10

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/853350-how-to-place-text-label-near-line-but-avoiding-other-lines-and-labels

评论： KAE 2023-11-14

I would like to plot many lines and put a text label near each line. But often the text labels lie over other lines or labels, and so can't be read. How can I reduce visual clutter by smarter label placement? I don't want to move the labels manually due to the large number of plots. Nor should the label text have a white background because it would obscure the plotted data. A plot legend doesn't work due to the large number of lines; the label needs to be near the line instead.

Here's an example of how the labels are hard to read. How can I clean this up?

figure;
x = 1:10;
nLine = 15;
y = rand(size(x));
for iLine = 1:nLine
    y = rand(size(x));
    h = plot(x, y); % Plot a line
    hold on; 
    
    % Put a text label at the max of the line
    iMax = find(y == max(y)); % Find the highest point
    % Label the line
    t = text(x(iMax), y(iMax), ['Line ' num2str(iLine)], 'color', get(h, 'color'));
    % Could also randomly pick a point on line to label; that will be
    % cluttered too
    
    % Instead, want to place label near line but where it won't overlap
    % anything
end
box off;
% Looks so cluttered!

3 个评论
显示 1更早的评论隐藏 1更早的评论

KAE 2021-6-10

Thanks, very useful! I do hope smart placement is possible because clabel places text in a smart way, if I recall. I suppose I could get the location of all existing objects in the plot and maximize a label's distance to them, while minimizing its distance to the points on the line that it's labeling.

dpb 2021-6-10

"I do hope smart placement is possible because clabel places text in a smart way..."

But not smart in the sense of your problem -- it has defined isolines in which it simply inserts a value every so often; it doesn't have to worry about another isoline intersecting or occluding the one it's currently working on (at least with smooth surfaces for which it was developed; I supposed one could dream up a counter-example).

If data are scattered all over as your example of random lines is; then KAE's suggestion of labeling at the ends of the lines or a vertical legend outside the axes is about the only choice there is.

It's not an easy problem in general, virtually all references to algorithms I found are for cartography applications that could be adapted I suppose, but in a fairly quick look I didn't find anything specific.

请先登录，再进行评论。

请先登录，再回答此问题。