How to ignore specific parts of a plot ??

Hello everybody,

I have the attached plot, the circles and the ellipse are holes so in my plot I want to leave blank inside the circles and the ellipse I do not want the lines cut through those.

Any suggestions how to do that??

[dpb]

How were they created? Is there data inside the boundaries; if so turn it to NaN and see if that's enough of a clue to the handle graphics to not display those locations.

[Ceren GURKAN]

Here how they are created:

if true
x = linspace(min(X_plot(:,1)), max(X_plot(:,1)), 100); 
y = linspace(min(X_plot(:,2)), max(X_plot(:,2)), 100);
[xg, yg] = meshgrid(x, y);
Fx = TriScatteredInterp(X_plot(:,1), X_plot(:,2), qx);
Fy = TriScatteredInterp(X_plot(:,1), X_plot(:,2), qy);
qxg = Fx(xg,yg);
qyg = Fy(xg,yg); 
figure(1); hold on; 
hl = streamslice(xg, yg, qxg, qyg);
set(hl, 'color', 'b','Linewidth',0.5);
end

It is actually quite tricky to remove the data corresponding to those holes this is why I am looking for another solution.

Answer 1

[Jason Nicholson]

Given the mesh I see in your picture, the following works becuase your elements are not distorted and your edges are very close to linear:

xg = linspace(min(X(:,1)), max(X(:,1)), 100);
yg = linspace(min(X(:,2)), max(X(:,2)), 100);
[xg, yg] = ndgrid(xg, yg);

elementNumber = tsearchn(X, T(:,1:3), [xg, yg]);
nans = isnan(elementNumber);
Fx = TriScatteredInterp(X_plot(:,1), X_plot(:,2), qx);
Fy = TriScatteredInterp(X_plot(:,1), X_plot(:,2), qy);
qxg = Fx(xg, yg);
qyg = Fy(xg, yg);
qxg(nans) = NaN;
qyg(nans) = NaN;
figure(1); hold on; 
hl = streamslice(xg, yg, qxg, qyg);
set(hl, 'color', 'b','Linewidth',0.5);

[Jason Nicholson]

If this works for you, please mark this as the correct answer. It helps my reputation points on MATLAB answers.

[Ceren GURKAN]

Jason thank you, your command is right my elements are not distorted. When you the above code though I get the following error message

   Error using ==> tsearchn at 27
   X and XI must have the same number of columns.

Since X has 2 columns and xi has 200 of them. How can I defeat this?

[Ceren GURKAN]

Jason, not exactly what you wrote but the following does me a bit better:

if true
x = linspace(min(X(:,1)), max(X(:,1)), 100);
y = linspace(min(X(:,2)), max(X(:,2)), 100);
xi=[x;y]';
elementNumber = tsearchn(X, T(:,1:3), xi);
[xg, yg] = meshgrid(x,y);
nans = isnan(elementNumber);
Fx = TriScatteredInterp(X(:,1), X(:,2), qnewx);
Fy = TriScatteredInterp(X(:,1), X(:,2), qnewy);
qxg = Fx(xg, yg);
qyg = Fy(xg, yg);
qxg(nans) = NaN;
qyg(nans) = NaN;
figure 
hl = streamslice(xg, yg, qxg, qyg);
set(hl, 'color', 'b','Linewidth',0.5);

end

How it looks is the following :

Thanks a lot.

[Jason Nicholson]

I can look at more tonight. Can you post the qx and qy vectors so that I can look closer at the problem?

Answer 2

[Jason Nicholson]

Can you post the data that made this plot?

My hunch is the main issue is TriScatteredInterp used Delaunay triangulation of your data and the holes were meshed in the TriScatteredInterp function. If instead you used the original triangulation to do the interpolation, then you should be able to get the desired result. If you can post the mesh, I can post code that does the interpolation without TriScatteredInterp. My code would rely on tsearchn to locate the element that contains point [xg, yg] and then use the local barycentric coordinates to interpolate. The key though is to use your mesh to define connectivity of your points. Points [xg, yg] that are not contained in an element will be returned as nan's and thus will not show up on the plot. So something like this:

[elementNumber, localElementCordinates] = tsearchn(xyCordinates, myTriangulation, [xg, yg]);
notNan = ~isnan(elementNumber);
qxg = nan(size(xg,1), 1);
qyg = nan(size(yg,1), 1);
% connectivity of elements containing points
elementsContainingPoints = myTriangulation(notNan,:);
% weights of nodes for given uq points that are contained in elements
weights = localElementCordinates(notNan,:) 
qxg(notNan) = sum(weights.*qx(elementsContainingPoints),2);
qyg(notNan) = sum(weights.*qy(elementsContainingPoints),2);

[Ceren GURKAN]

Jason,

I was thinking I have the problem because of simply creating a mesh grid with x and y that is created by linspace through all the mesh (ignoring the holes I have). Guess this indirectly causes what you have commented. Anyway here I added the X (coordinates) and T(connectivities) mat files, I will try your suggestion and commend here what I got.

Thanks.

[Jason Nicholson]

Okay so may answer may not work because your mesh has 6 node parabolic triangles rather than 3 node linear triangles.

Really you need a set of tools for parabolic triangles: locate an element containing a given point barycentric coordinates (local weighted coordinates of the nodes).

Do understand what TriScatteredInterp is doing when you use natural neighbor interpolation?

I'll have to think about this a little more.

[Jason Nicholson]

How are you generating your mesh?