how to create a matrix of values from a X and Y coordinates

Hi

I have a vector u (dots) with its x,y coordinates. Each vector has the same size. I wanto to create a matrix u using x,y coordinates. How I can do it, taking into account that x and y don't form a perfect rectangle.? I mean, there are empty values of u without their coordiantes x,y respectively as you can see in the figure. Empty values can be filled by NaN.

Answer 1

[Jorge Peñaloza Giraldo]

Thank you everyone for you help.

I all ready solve using the following code.

d = [2.0000         0    0.3800;
     2.0000    5.0000    0.3480;
     2.0000   25.0000    0.3280;
     3.0000    2.0000    0.3190;
     3.0000   50.0000    0.2120;
     5.0000   10.0000    0.1380;
     5.0000   60.0000    0.1100;
     5.0000  100.0000    0.0870;
     7.0000    5.0000    0.0690;
     7.0000    9.0000    0.4113;
     7.0000   45.0000    0.3807;
     7.0000   95.0000    0.3678;
     9.0000   10.0000    0.3207;
     9.0000   35.0000    0.1765;
    15.0000   65.0000    0.1123;
    15.0000   75.0000    0.0917;
    15.0000   99.0000    0.0727;
    15.0000  125.0000    0.0573;
    15.0000  150.0000    0.3811];
X = d(:,1); Y = d(:,2); Z = d(:,3);
Xs = unique(X);
Ys = unique(Y);
Xi = arrayfun( @(x) find(Xs==x), X );
Yi = arrayfun( @(y) find(Ys==y), Y );
Li = Yi + (Xi-1) * numel(Ys);
XYZ = nan(numel(Ys), numel(Xs));
XYZ( Li ) = Z;

I found it in this link: https://www.mathworks.com/matlabcentral/answers/38731-how-to-create-a-regular-data-matrix-from-irregularly-sampled-xyz-data-with-different-x-and-y-coordin

Answer 2

[Shunichi Kusano]

First, the xy cordinate must be converted to indices which are the coordinates of your matrix.

% dx, dy: minimum sampling interval of your x, y coordinates.
xi = x / dx;
xi = xi - min(xi) + 1; % shift so that the minimum index is 1.
yi = y / dy;
yi = yi - min(yi) + 1; 

Then, matrix size is decided from max(xi) and max(yi).

u_mat = NaN(max(yi), max(xi));

Finally, the value u is stored in u_mat. Before that, the indices of xi and yi is converted into linear index.

lin_i = sub2ind(size(u_mat), yi, xi);
u_mat(lin_i) = u;

Some modification may be needed. hope this helps.

[Jorge Peñaloza Giraldo]

Thank you for your answer.

I have a problem with dy beacuse the minimum is cero. There are many coordinates that are repeated. Do you have another idea? Also my dx and dy are not iinteger numbers. Therefore, creating a matrix using NaN does not work.

[Shunichi Kusano]

for the problem on dy, please remove the repetetion by unique command. Then, dy will be correctly computed:

dy = min(diff(unique(y)))

But, if your xi and yi include non-integer (or your data cannnot be on a uniform grid), the proposed procedure does not work as you mentioned.

By the way, it seems that you can avoid interpolation with griddata. Just replace interpolated values to NaN at the point where you don't have u.

I'm sorry, I have no more idea so far.

Answer 3

[Steven Lord]

Make a regular grid of Xq and Yq coordinates with meshgrid then pass your scattered data and the grid coordinates to the griddata function.

[Jorge Peñaloza Giraldo]

Tha idea is not interpolate the values. Also coordinates x and y should match with u.