problem with scatteredInterpolant: are there any limits?

the xyz data file consists out of 3157394 data triples like this:

417826.974 5333045.048 1636.000

417826.974 5333045.128 1682.000

417826.974 5333045.208 1744.000

417826.974 5333045.288 1753.000

417826.974 5333045.368 1674.000

417826.974 5333045.448 1689.000

with the key data:

min(x) = 417740; max(x) = 417870; min(y) = 4177412; max(y)= 5333100; min(z)= 0; max(z) = 11054;

length(x) = length(y) = length(z) = 3157394;

First, everything works fine using:

TT=readtable(FileNameNeu,'PreserveVariableNames',true,'NumHeaderLines',1);

x=table2array(TT(:,1));
y=table2array(TT(:,2));
z=table2array(TT(:,3));

scatter3(x,y,z,4,z,'.'); view(2); grid on; daspect([1 1 1]);

But when I try this:

F = scatteredInterpolant(x,y,z,'natural','none'); :
[X,Y] = meshgrid(min(x):0.5:max(x),min(y):0.5:max(y)); 
Z = F(X,Y); 

then I receive the error:

Error using scatteredInterpolant
The coordinates of the input points must be finite values; Inf and NaN are not
permitted.

Error in FillGaps (line 44)
F = scatteredInterpolant(x,y,z,'natural','none'); % create the interpolant

But...there are no Inf's and NaN's in the data file.

The same occurs if I reduce the values of x and z using:

x=x-min(x); y=y-min(y);

What I am doing wrong? Thanks a lot,

Harry

[John D'Errico]

It would so much help if you actually provided the data. Shifting the data, and still having a problem tells me the problem is not a numerical problem. However, it is still possible you have something inf or NaN in there, and you do not realize it.

[Bruno Luong]

The data cannot have -Inf or Inf since min and max are finites. However NaN is possible.

[Steven Lord]

Since you're using release R2022b you can use the anynan or allfinite functions introduced in release R2022a to determine which of the variables contain a non-finite value.

[Harald von der Osten]

thanks a lot to all for your help. Finally Steven Lord's hint on anynan and allfinite helped a lot. Every some thousands line there was a text like "line xyz" in the datafile. Crazy.

[Harald von der Osten]

  417768.334  5333147.288        0.000
  417768.334  5333147.368        0.000
  417768.334  5333147.448        0.000
line    4177684
  417768.414  5332990.328        0.000
  417768.414  5332990.408        0.000
  417768.414  5332990.488        0.000
  417768.414  5332990.568        0.000

Answer 1

[Star Strider]

It would help to have the file.

It appears to me that the data might be gridded. To determine that, the easiest way would be:

TT=readtable(FileNameNeu, 'VariableNamingRule','preserve')

x = TT{:,1};

y = TT{:,2};

z = TT{:,3};

[Ux,ixx] = unique(x)

RowSize = unique(ixx)

X = reshape(x,RowSize,[]);

Y = reshape(y,RowSize,[]);

Z = reshape(z,RowSize,[]);

figure

surf(X, Y, Z)

grid on

xlabel('X')

ylabel('Y')

zlabel('Z')

If that fails (the data not being gridded appropriately), try this —

L = size(TT,1);

xv = linspace(min(x), max(x), L);

yv = linspace(min(y), max(y), L);

[X,Y] = ndgrid(xv, yv);

Z = griddata(x, y, z, X, Y);

figure

surf(X, Y, Z)

grid on

xlabel('X')

ylabel('Y')

zlabel('Z')

One of those should work.

EDIT — (13 Nov 2022 at 23:52)

Corrected typographical errors.

.

[Harald von der Osten]

thanks a lot for your efford. Finally, the error occured because of the text "line xyz" in the datafile....

[Star Strider]

As always, my pleasure!

Answer 2

[Bruno Luong]

You'll have problem anyway since your data is not centered and especially not normalize.

x range is 130 and y range is 1155688, almost 10000 larger than xrange. Scattered interpolation is based on Delauray triangulation, and that alone will give garbage out.

I believe it written somewhere (tip) this warning in the doc page.

So please do CENTERING and NORMALIZATION data as preproceesing step before interpoltion.