griddata (averaging instead of interpolation)

I have many raw data points of a pipe surface that i scanned with a 3D scanner. the amount of points is about 4 times as much as my intended grid points. I would like to average the data (about a small cell) at a every specific grid point (to lessen the scanner noise), instead of using matlab 'linear' 'cubic' or 'nearest neighbor'.

How can i do so, is there better function than griddata that i can use?

thank you

Answer 1

[Bruno Luong]

As I understood you have 2D data r depending of theta and z (and not 3D).

theta = rand(100000,1)*2*pi;
z = rand(100000,1)*4-2;
r = 1+theta.*exp(-theta.^2-z.^2);
r = r+0.05*randn(size(r));

thetai=linspace(0,2*pi,33);
zi = linspace(-2,2,33);

% FEX: http://www.mathworks.com/matlabcentral/fileexchange/23897-n-dimensional-histogram
tz = [theta(:) z(:)];
[~, ~, ~, loc] = histcn(tz, thetai, zi);
r_avg = accumarray(loc, r(:), [length(thetai) length(zi)]) ./ ...
        accumarray(loc, 1, [length(thetai) length(zi)]);

figure(1);
subplot(1,2,1);
plot3(z,theta,r,'.');
subplot(1,2,2);
surf(zi,thetai,r_avg)

Bruno

[Thuan Le]

[~,~,~,loc] = histcn(tz, thetai, zi);

im getting error Expression or statement is incorrect--possibly unbalanced (, {, or [.

i downloaded the histcn function.

[Bruno Luong]

The reason is your matlab is old and does not support ~ operator, replace with:

[trash,trash,trash,loc] = histcn(tz, thetai, zi);

[Thuan Le]

your are correct, my matlab is too old. I'm also looking into TriScatteredInterp, using the natural neighbor interpolation. Anyway thank you for your help.

Answer 2

[John D'Errico]

No, this will not explicitly average your data points in each bin. It does better than that.

http://www.mathworks.com/matlabcentral/fileexchange/8998-surface-fitting-using-gridfit

[Thuan Le]

gridfit is not quite what i was looking for. it was too bias toward smoothness for my application. Im trying to capture the scallops and wear on the pipe surface. averaging the data about a small cell should be best.

Any other suggestions?

Answer 3

[Thuan Le]

thank you for your quick reply.

i will take a look at the files and get back soon

Answer 4

[Thuan Le]

gridfit is not quite what i was looking for. it was too bias toward smoothness for my application. Im trying to capture the scallops and wear on the pipe surface. averaging the data about a small cell should be best.

Any other suggestions?

[Bruno Luong]

Please explain the "small cell". Is it a cell from prescribed 3D grid or something else?

[Thuan Le]

i have the raw data of (theta, z and r) which map out a surface of a pipe (alot of data point). im specifying a grid of theta and z, so that i can interpolate the radius. so let say for theta = 5 (+-0.5) , z = 6 (+-0.5 to form a 'cell'), interpolate radius by averaging any r values in that theta = 5+-0.5, z = 6+-0.5 range.

[John D'Errico]

You CAN control the amount of smoothing using gridfit. All you have shown is that you did not read the help.

[Thuan Le]

and where is this `help`files that you talked about in your `multiple`posts? all i can see is understanding gridfit.doc and gridfit.pdf