On 27 Giu, 19:51, PeterOut <MajorSetb...@[EMAIL PROTECTED]
> wrote:
> On Jun 27, 1:25 pm, Martin Leese <ple...@[EMAIL PROTECTED]
>
> wrote:
>
> > PeterOut wrote:
> > > If you have a set of points belonging to 2 or more intersecting
lines
> > > in 2D, what is the best way to fit the lines to the points. Would
it
> > > be the Hough transform?
>
> > Is this a class assignment? If so, say so.
> > If not, describe your problem (where do the
> > points come from).
>
> > --
> > Regards,
> > Martin Leese
> > E-mail: ple...@[EMAIL PROTECTED]
> > Web:http://members.tripod.com/martin_leese/
>
> No. I finished with university cl***** 14 years ago. For my
> particular problem, there are a number of reasons why I may not be
> more specific about the source. I can say that they are fairly clean
> (as opposed to random) points on lines that intersect. I would
> suspect the number of such lines is about 10 and there are roughly
> 5-10 points for each line.
I assume you are speaking of _straight_ lines.
I think you could use the Hough transform, with distance and angle as
parameters, to separate the points belonging to each line and get a
first approximation of the parameters.
Then if you need more precision you should fit straight lines to the
subsets you got. See eg David Alciatore, Rick Miranda, The Best Least-
Squares Line Fit, in Graphics Gems V for a method which minimizes the
normal squared distances.
You may also want to use a statistical method like Least Median of
Squares to get rid of possible outliers.
|
