3D Interpolation of randomly distributed anchor points

[NKane]

Hi all,

I've got a problem with analysing some 3D-data with LabVIEW. Maybe you can give me a hint for the right path.

What I want to do:

I've got a 3D color map built by 61 anchor points (X, Y) and the measured values (Z). The anchor points are unequally spaced and distributed over a round surface (see picture 1 and 2). I want to interpolate the map with an appropiate interpolation algorithm (spline?) to

a) smooth the map

b) get more points for analysis.

For analysis I need the minimum and maximum of the interpolated map. If I can calculate these values the interpolation isn't important anymore.

My first attempt to solve this problem was to create a close meshed grid of XY values and use the 3D polynomial fit. This new grid must also have the old 61 grid points as elements at the correct place. Therefore the new grid must be so narrow that is has lots of points and one calculation of map lasts around 10 minutes. Thats too long.

Thanks in advance for your help.

Greets

Kane

[altenbach]

If you have LabVIEW 8.2.1 or later, you can use the mathscript griddata function.

See http://forums.ni.com/ni/board/message?board.id=170&view=by_date_ascending&message.id=249724#M249724

If you want to fit it to a polynomial plane of a desired order to yuor points, you could use "general LS fit". Here's an example (ignore the levenberg marquardt parts).

http://forums.ni.com/ni/board/message?board.id=170&view=by_date_ascending&message.id=129944#M129944

Attach some typical data and I'll give it a try. 🙂

Message Edited by altenbach on 12-17-2007 09:21 AM

[NKane]

Hi altenbach,

first of all thanks for spending time on this problem. As a matter of fact i played around with your impressive 3D interpolation VIs already, but I didn't get reasonable results (caused by wrong usage of the operator called Kane, i guess 😉 ). I attach some example data file (ASCII) to this reply. First column are the X values, second the Y values, third the Z values. Maybe you've more success.

Thanks in advance.

Kane

Message Edited by NKane on 12-18-2007 05:22 AM

[GerdW]

Hi Kane,

attached is a pic generated from your sample data using mathscript "griddata" function... And a picture from the vi using this very function

[NKane]

Hi Gerd,

i give your code a try and get some reasonable results, but not quite the one i wanted. First of all I get a rectangular grid instead of an elliptic one as shown in your picture (altenbachs link does the same). I also don't understand some wires and structures in your code example (what are the other cases of the case structure, for example?).

So could you provide the VI so that I can look by myself what it is doing? And do you have a hint for the elliptic grid?

Thanks a lot in advance.

Greets

Kane

[GerdW]

Hi Kane,

the code snippet is from a bigger program designed for visualization of all kind of wafer measurements (silicon wafers for sensor R&D)  - hence the circular picture!

The other cases are doing approximations using different methods (bilinear, several distance weighting algorithms) to get a similar picture.
They all result in an rectangular shaped grid which is cutted to wafer dimensions or chip mask design while drawing the picture. I just do a comparison of chip center with wafer circumference to decide between "drawing" and "no drawing" to get a circular shape... (see attached pic for user interface)

[NKane]

Ah, allright.

I've only one more question to you. Where can i vote for answers? I'm to blind.

Greets

Kane

[GerdW]

Hi Kane,

you can "vote" in the lower right part of the message - but from your status bar color (=all white) I think you're not allowed to vote so far...