Griddata - extrapolation beyond the Delaunay triangulation

Hi all!

I have an array (for example):

A = [1114.74,419.09,139.578;
    1102.49,442.43,139.188;
    1081.19,433.00,138.894;
    1094.47,414.38,139.072;
    1070.44,408.88,138.810;
    1068.03,369.45,138.715;
    1080.25,392.28,138.941;
    1085.77,374.79,139.178];

And wrote a code

[griNodN,griNodE] = meshgrid(min(A(:,1)):0.1:max(A(:,1)),min(A(:,2)):0.1:max(A(:,2)));
valGridDef = griddata(A(:,2),A(:,1),A(:,3)',griNodE,griNodN,'linear');
isolin=(min(A(:,3)):0.05:max(A(:,3)));
clim([min(isolin) max(isolin)]);
contourf(griNodE,griNodN,valGridDef,isolin);
colorbar
plot(A(:,2),A(:,1),'rs','MarkerSize',10,'LineWidth',3)

the result looks as expected

In documentation is noted "For all interpolation methods other than 'v4', the output vq contains NaN values for query points outside the convex hull of the sample data. The 'v4' method performs the same calculation for all points regardless of location."

But v4 gives irrelevant result.

Is there another way to at least approximately extrapolate the surface beyond the Delaunay triangulation?

Answer 1

[Mathieu NOE]

hello

to have access to extrapolation , use scatteredInterpolant instead of griddata

A = [1114.74,419.09,139.578;
1102.49,442.43,139.188;
1081.19,433.00,138.894;
1094.47,414.38,139.072;
1070.44,408.88,138.810;
1068.03,369.45,138.715;
1080.25,392.28,138.941;
1085.77,374.79,139.178];

[griNodN,griNodE] = meshgrid(min(A(:,1)):0.1:max(A(:,1)),min(A(:,2)):0.1:max(A(:,2)));

F1 = scatteredInterpolant(A(:,2),A(:,1),A(:,3),'linear','linear');
valGridDef = F1(griNodE,griNodN);
isolin=(min(A(:,3)):0.05:max(A(:,3)));
contourf(griNodE,griNodN,valGridDef,isolin)
hold on
colorbar
plot3(A(:,2),A(:,1),A(:,3),'rs','MarkerSize',10,'LineWidth',3)

[Andrey Melnikov]

Great! Thanks a lot

[Mathieu NOE]

my pleasure !

do you mind accepting my answer ? tx