Discussion: View Thread

  • 1.  Polygon descriptors

    Posted 10-24-2011 10:41

    Hello All!

    This is my first time attempting to describe polygons and I am a little confused. 

    Each sample in datasets has "n"-points in (x,y), cartesian coordinate and I want to estimate density function for n-points. Since it is high-dim problem depending on "n" and our interest is to estimate the probability of the event with "n-points", I connect n points and make a n-polygon to reduce dimensionality. 

    So, I am trying to find polygon descriptors to characterize/classify polygons when the number of edge(vertex) is fixed. So far, area and perimeter of polygons are considered, but I think it is not enough for the difference by rotation. 

    What I hope to do is to find few variables to explain polygons, rotation-sensitive. 

    I would really appreciate your comments and suggestions.

    Thanks.

    Mina    


    -------------------------------------------
    Mina Yoo
    PSU
    -------------------------------------------


  • 2.  RE:Polygon descriptors

    Posted 10-24-2011 13:19

    New to me too, but here are some suggestions --
    1.  Ratio of area to perimeter.
    2.  Length of projection on x-axis.
    3.  Length of Projection on y-axis.

    Good luck.

    -------------------------------------------
    Charles Mann
    Charles R Mann Associates Inc
    -------------------------------------------








  • 3.  RE:Polygon descriptors

    Posted 10-25-2011 17:32
    I don't understand what makes it a high-dim problem depending on "n".  Each point has a (x,y) pair; right?  Does that not make it 2-dim no matter how big "n" is?

    -------------------------------------------
    Eric Siegel
    Biostatistician
    Univ of Arkansas for Medical Sciences
    -------------------------------------------