Ellipse fitting in python?

[Ron Kneusel]

Is there a python package out there that supports fitting a set of 2D
points to an ellipse? I haven't been able to find one.

Thanks!

Ron

 

[John J. Lee]

Is there a python package out there that supports fitting a set of 2D
points to an ellipse? I haven't been able to find one.

This is a set of points with some noise (scatter), right?

A not-very-thoughtful answer: a sledgehammer method is to throw an
implementation of a standard nonlinear least-squares algorithm at the
problem. For example, SciPy contains a wrapper of MINPACK's nonlinear
LS solver . There are several implementations of the same
(Levenburg-Marquardt) algorithm in pure Python floating around, too.

If there's no scatter, it's time to dig out your old algebra textbooks



John

 

[Fernando Perez]

This is a set of points with some noise (scatter), right?

A not-very-thoughtful answer: a sledgehammer method is to throw an
implementation of a standard nonlinear least-squares algorithm at the
problem. For example, SciPy contains a wrapper of MINPACK's nonlinear
LS solver . There are several implementations of the same
(Levenburg-Marquardt) algorithm in pure Python floating around, too.

A quick comment in case it's not obvious: an ellipse is not a function in
Cartesian coordinates, and most standard LS solvers expect a function. So
it's probably best to convert things over to polar before feeding the data to
the LS solver.

f

 

[Helmut Jarausch]

Is there a python package out there that supports fitting a set of 2D
points to an ellipse? I haven't been able to find one.

It's not so easy; it's a nonlinear orthogonal distance regression problem.
I know of a FORTRAN package and I myself have a (more general) package written
in Scilab ( a Matlab "clone" )

If you have just a few datasets, I'll compute a fit for you.
But implementing it in Python would be much work.


--
Helmut Jarausch

Lehrstuhl fuer Numerische Mathematik
RWTH - Aachen University
D 52056 Aachen, Germany

 

[Andreas Lobinger]

Aloha,

Is there a python package out there that supports fitting a set of 2D
points to an ellipse? I haven't been able to find one.

See http://www.magic-software.com/Approximation.html
Here are C++ Sources, you will need a interface like SWIG for using this
in python.

Wishing a happy day
LOBI

 

[Christopher T King]

A quick comment in case it's not obvious: an ellipse is not a function in
Cartesian coordinates, and most standard LS solvers expect a function. So
it's probably best to convert things over to polar before feeding the data to
the LS solver.

You could break both the ellipse and the data into two pieces (say, y>=K
and y<K). Polar coordinates might get you more uniform results, though.

 

[Robert Kern]

It's not so easy; it's a nonlinear orthogonal distance regression problem.
I know of a FORTRAN package and I myself have a (more general) package
written
in Scilab ( a Matlab "clone" )

I wrote a wrapper to the FORTRAN ODRPACK.

http://starship.python.net/crew/kernr/source/odr-0.6.tar.gz