Plot contours from coordinates as a line

Hey there, I'm struggling with a problem regarding plotting a shape directly from x,y coordinates.

I have a set of (x,y) coordinates of points that describes several contours in the (x, y) plane.

I can display those contours using :

scatter(x, y)

The problem using this function is that the figure is quite heavy, and exporting it to pdf for, say, editing it on Inkscape results in a vector image with a weight of more than 10MB that Inkscape has troubles handling. Note that the coordinates contain more than 2000 (x,y) pairs.

The solution would be to plot those contours as a line rather with discrete points, but I don't find a way.

The two issues I get that make impossible to use plot rather than scatter is that :

• coordinates are in unordered manner so it gives me multiple lines going everywhere
• coordinates reprensent multiple contours at once so the different regions are connected with a line at some point.

I guess I could achieve something with the contour or surf functions, setting a fake z dimension that would describe the contours but I didn't manage...

Any insights? You'll find attached an example of two objects whose contours are described by their x,y coordinates.

Cheers,

Guillaume

[KSSV]

Try boundary function.

[Guillaume Le Goc]

I didn't know this function, it could have worked but it simplifiestoo much some complex shape and doesn't really work when there is two objects.

Answer 1

[darova]

Solution

• separate each data using logical indexing
• order data using angle

clc,clear

s = load('xycontours.txt');
x = s(:,1);
y = s(:,2);
ix = x < 0.25;
x1 = x(ix);             % left points
y1 = y(ix);
x2 = x(~ix);            % right points
y2 = y(~ix);
    % center each data set and convert to polar (calculate angle)
[t1,r1] = cart2pol(x1-mean(x1),y1-mean(y1));
[t2,r2] = cart2pol(x2-mean(x2),y2-mean(y2));
[~,ix1] = sort(t1);     % sort angle
[~,ix2] = sort(t2);     % sort angle

plot(x1(ix1),y1(ix1),'r')
line(x2(ix2),y2(ix2))

[Guillaume Le Goc]

That's clever, nice one!

But unfortunately I have several sets of coordinates such as this one, and I wanted an unsupervised way (ie. with no knowledge of number of regions, where to threshold...). I guess it's feasible to build something that automates this but I don't really have time and it's kinda okay the way it is with scatter. I thought maybe MATLAB could do this "natively".

[darova]

Do you know how many datasets you have? Maybe clusterdata

s = load('xycontours.txt');
x = s(:,1);
y = s(:,2);
ix = clusterdata(s,'maxclust',2);   % number of data sets

for i = 1:numel(unique(ix))
    x1 = x(ix==i);                  % choose dataset
    y1 = y(ix==i);
    [t1,r1] = cart2pol(x1-mean(x1),y1-mean(y1));
    [~,ix1] = sort(t1);             % sort angle
    line(x1(ix1),y1(ix1))
end