Linear Clustering Methods?

[capucsc]

Hi,

Are there any clustering methods well suited for detecting linear clusters of points? I'm working with an array of (X,Y) data that form several lines. I've attached a picture of an example dataset. I would like to identify each line, isolate the points associated with the line and do a linear fit to extract the slope. I've explored LV clutering tutorials, but every method I try isolates circular clusters of points. If clustering doesn't work, are there any other ways  to extract these data? 

Thanks, 
Chris 

[Darin.K]

Have you tried Gaussian Mixture Modeling?  It sounds like you have been using K-Means methods which tend to have trouble with high aspect ratio regions since they use simple distance assignments.  Usually you end up having to generate a large number of regions to get a decent result, and your lines wind up being covered with many circular subregions.  I am sure that there must be some methods in the statistics literature, just brush up on your hyperplanes before you start.