And the first hit even has the words "plane fitting" in it.

So if they don't work, then you should be more explicit about the "why" and the "what".

If you know that your plane will be close to horizontal (or can be reliably pre-transformed to make it so), then the Hough transform is a viably efficient approach.

If you gave a description of how the points are gathered, or could post a (not too large) 'typical' dataset, then you might get better leads.

But the more generalised an approach you need--ie. the less pre-constraints you can impose on the problem--the harder it is to solve efficiently. A fully generalised solution can be prohibitively expensive.

You may try to use Math::GSL::Multifit which is a wrapper over the multi-dimensional least square fitting functions in GSL. You need the linear fit functions of course.

Thanks Folks!
I can't get my head round Statistics::Regression, can you simply input 3D cartesian co-ordinates and output an equation for the plane?
Does anyone have an example please?

I have to admit the documentation doesn't tell you how to do what the module claims. Statistics::LineFit, on the other hand, does have the least squares methods required to extend a line of regression analysis into 3D, provided you can cope with the required matrix iterations, see http://www.weibull.com/DOEWeb/estimating_regression_models_using_least_squares.htm

One world, one people