Re: Best Fit Plane - Parallel to xyplane!


p4pradeep@xxxxxxxxxxx wrote:
Thanks for the reply.

We have already considered this solution to take the average of all z
coordinates, but this was not accepted. Actually what we need is that,
we want a plane which is the mean of "all the points" on the whole
surface (not only the z points we have got, but also all the points on
the surface).

But, if you only give the algorithm the coordinates of the points, all
it has to work with is... the coordinates of the points!


Suppose we have four z points. If we take mean we get the mean of only
of these four points. But what we are thinking is, if you create a
surface using these four points

There will be many (understatement) surfaces that contain those four
points - which do you want to choose?

, we also get multiple z points on this
surface and the mean of all these z points wont be the same as the mean
of just four points we considered before, right? May be my idea is
completely wrong :-S Could you help?

I suspect that in your particular application, all your surfaces are in
some sense 'the same shape', so that when you think of four points, it
is easy to visualize 'the' surface that contains them. However, I say
again, if all we have is the points, all we can use is the points.

If there is some reliable (that is to say, algorithmic) way of
determining 'the' surface from the points, then we are in business.
Otherwise, we need a better (larger and representative) sample of
points to get a better fit to the surface.

--
Larry Lard
Replies to group please

.

Relevant Pages

  • Re: qubits
    ... >>the hardware level, I suspect we would need some rather drastic advances ... B-) My point is that I suspect a port of existing 3GLs ... For example, if the algorithm utilizes ... before implementing any algorithm /effectively/ on a Qubit machine ...
    (comp.object)
  • Re: FFT algoritms
    ... I suspect that Windows version is dependent on using IE. ... they have an algorithm that works without bit reversal sorting. ... square waves instead of sinusoids as my basis functions. ... algorithm depends on the fact that sinusoidal basic vectors ...
    (comp.dsp)
  • Re: How can I compute the scan rate (sr) ???
    ... I suspect the algorithm is correct only when freemem < ... the real vmstat isn't doing any such calculation. ... kernel is doing the calcuation. ...
    (comp.unix.solaris)
  • Re: FFT algoritms
    ... I suspect that Windows version is dependent on using IE. ... they have an algorithm that works without bit reversal sorting. ...
    (comp.dsp)
  • Re: FFT algoritms
    ... I suspect that Windows version is dependent on using IE. I'll see if local library can get me a hard copy on inter-library loan. ... they have an algorithm that works without bit reversal sorting. ... I'm not interested in the computation time but in whether or not I can follow what's going on. ...
    (comp.dsp)
Quantcast