Effacer les filtres
Effacer les filtres

reconstruct matrix from vector coordinates

1 vue (au cours des 30 derniers jours)
Nicolas
Nicolas le 24 Jan 2017
Commenté : Walter Roberson le 25 Jan 2017
Hello, I have 3 vectors (x,y,data) each equals 1x6358. How can I recreate the matrix DATA based on the indices x and y? Thank you

Réponse acceptée

Walter Roberson
Walter Roberson le 25 Jan 2017
Possibly you should be using scatteredInterpolant() or TriScatteredInterp to create an interpolant that you would then sample at points on a 2D grid.
Or possibly you should use
[xu, ~, xuidx] = unique(x);
[yu, ~, yuidx] = unique(y);
nxu = length(xu);
nyu = length(yu);
DATA = accumarray( [xuidx(:), yuidx(:)], data(:) );
This would be primarily for the situation where your x and y values form a grid but not necessarily of integer values and not necessarily equi-distant. This particular form of the code expects that all the x values that are intended to be treated as equal are exactly equal -- for example although 1 and 1-eps both display as 1, they are not exactly equal and this code would treat them as different. (If you do have a grid but the values are not exactly equal then there are ways to deal with that, such as by using uniquetol())
  2 commentaires
Nicolas
Nicolas le 25 Jan 2017
Yes, the mesh is more compact in some areas than some others.
Walter Roberson
Walter Roberson le 25 Jan 2017
The nxu and nyu lines above are not needed; I forgot to edit those out. You can get away with
[xu, ~, xuidx] = unique(x);
[yu, ~, yuidx] = unique(y);
DATA = accumarray( [xuidx(:), yuidx(:)], data(:) );
or, if needed, uniquetol() instead of unique()
This would produce a grid that would not necessarily be equally spaced. The vectors xu and yu give the actual coordinates.
If you wanted to create an equally spaced mesh, then probably after the above you would use griddedInterpolant together with a mesh of the points you wanted to sample at.

Connectez-vous pour commenter.

Plus de réponses (1)

Stephen23
Stephen23 le 24 Jan 2017
Modifié(e) : Stephen23 le 24 Jan 2017
You could use sub2ind:
sz = [max(x),max(y)];
ix = sub2ind(sz,x,y);
mat = NaN(sz);
mat(ix) = data;
  3 commentaires
Walter Roberson
Walter Roberson le 25 Jan 2017
Your x and y are not indices in the MATLAB sense.
Nicolas
Nicolas le 25 Jan 2017
no they are spatial coordinates from the mesh of a numerical model. do I need to find a way to create the indices from the mesh in spatial coordinates?

Connectez-vous pour commenter.

Catégories

En savoir plus sur Computational Geometry dans Help Center et File Exchange

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by