Thanks for the appreciation and your link.
But I must say that this is not centroid in the strictest sense.

What I have done is to find a latitude which lies between the latitudes of
the vertices of the polygon
and a longitude which lies between the longitudes of the vertices of the
polygon.

This is clearly not the centroid of the polygon in spherical geometry, as I
discover it now.
However this serves my purpose in my project, though the real centroid would
have been the much desired utopia.

I will try to work on this problem of finding the exact centroid of a
spherical polygon in my spare time. I have few ideas in my mind which I want
to test. Thanks once again.


--
regards,


~Apurv Verma
   B. Tech.(CSE)
   IIT- Ropar
   Mobile - 09256587545






On Mon, Aug 1, 2011 at 8:11 AM, Michael Bedward
<[email protected]>wrote:

> Hi Apurv,
>
> Many thanks for the link to your code. I've only had a quick scan so
> far (sorry - busy with other things at the moment) but if I have any
> useful comments to offer when I look at it properly, I'll post them
> here.
>
> Meanwhile, have you seen this polygon-based approach ? (approach
> described in last third of doc)  It includes some error comparisons
> between lat/lon and projected coordinate centroids.
>
> http://www.jennessent.com/downloads/Graphics_Shapes_Online.pdf
>
> Michael
>
>
> On 31 July 2011 09:25, Apurv Verma <[email protected]> wrote:
> > Hi all,
> > Here is a method to calculate a centroid which lies inside a region.
> > Well I seem to have finished this coding.
> > I have shared a link to the class.
> > You just need to invoke the public method findMeanPosition(<the vector of
> > longitudes>), This will give you the mean longitude.
> > Calculating the mean latitude is easy, it is just a normal mean because
> > latitude values range from -90 to 90 so they don't bother us about caring
> > about crossing the 180.
> >
> > At present there are a lot of print statements, which I will remove once
> I
> > am done with writing the tests for this class.
> >
> >
> https://github.com/dapurv5/phyloGeoRef/blob/master/phyloGeoRef/src/nescent/phylogeoref/processor/utility/ComputeUtility.java
> > This is an O (n log n) algorithm where n is the number of points of the
> > polygon. Atleast theoretically it works out to be it.
> >
> >
> > --
> > thanks and regards,
> >
> > ~Apurv Verma
> >    B. Tech.(CSE)
> >    IIT- Ropar
> >    Mobile - 09256587545
> >
> >
> >
> >
> >
> >
> > On Tue, Jul 26, 2011 at 7:55 AM, Michael Bedward <
> [email protected]>
> > wrote:
> >>
> >> If the data points span most of the globe then the definition of
> >> inside-ness becomes a little difficult :)
> >>
> >> A link to your algorithm posted to the list would be much appreciated.
> >> It could well be useful for other applications.
> >>
> >> Good luck with it.
> >> Michael
> >>
> >> On 26 July 2011 00:44, Apurv Verma <[email protected]> wrote:
> >> > Hi Mike,
> >> > Yes on the plane it's easy. But in spherical geometry it's not more so
> >> > when
> >> > the points may span the entire globe, I have a previous algorithm but
> >> > just
> >> > wanted to know if there is something ready made already.
> >> > Once I implement that algorithm I will post a link to it on the list
> >> > here so
> >> > that people interested in calculating centroids on the surface of
> earth
> >> > might use it.
> >> >
> >> > Sure I will look at R.
> >> >
> >> > You are really doing a wonderful job, I see you handle a lot of emails
> >> > everyday. :P
> >> >
> >> > ~Apurv Verma
> >> >    B. Tech.(CSE)
> >> >    IIT- Ropar
> >> >    Mobile - 09256587545
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > On Mon, Jul 25, 2011 at 6:20 PM, Michael Bedward
> >> > <[email protected]>
> >> > wrote:
> >> >>
> >> >> Hi Apurv,
> >> >>
> >> >> Thanks for that - I understand better now.
> >> >>
> >> >> I guess by multidimensional scaling you mean any heuristic that would
> >> >> give a "good enough" solution with some indication of goodness of
> fit,
> >> >> e.g. classic MDS but also simulated annealing etc. I don't think
> >> >> either GeoTools, or the JTS library which it uses for geometry
> >> >> operations, have what you want, though I'd be happy to be corrected
> by
> >> >> someone else here.
> >> >>
> >> >> In the plane I **think** that the coordinate that minimizes the sum
> of
> >> >> distances to all points in a set is the median, ie. centroid.x =
> >> >> median of point.xs, centroid.y = median of point.ys. If your points
> >> >> are often close together (local or regional scales) you could just
> >> >> work with Cartesian coordinates in some convenient map projection.
> >> >> However, you emphasize spheroidal distance in your post so I'm
> >> >> guessing that your point sets cover a large area. In that case I've
> no
> >> >> idea whether there is an easy solution, but once again it depends on
> >> >> how critical the minimum sum of distances criterion is for your
> >> >> application.
> >> >>
> >> >> GeoTools does have the GeodeticCalculator class which can compute
> >> >> distances between points on a spheroid. So you could use that in
> >> >> conjunction with an optimizing algorithm from another library.
> >> >>
> >> >> As an aside, why are you searching in GeoTools for this ?  I would
> >> >> have thought the first stop would have been R (which has many
> packages
> >> >> for spatial analysis) or similar.
> >> >>
> >> >> Sorry I can't be more directly helpful but please let us know how you
> >> >> go.
> >> >>
> >> >> Michael
> >> >>
> >> >>
> >> >> On 25 July 2011 19:32, Apurv Verma <[email protected]> wrote:
> >> >> > Hi Michael,
> >> >> >  I am extremely sorry that I did not cc my message to the list last
> >> >> > time.
> >> >> > Thanks for providing me with the link.
> >> >> >
> >> >> > Well let me tell me you my problem in detail.
> >> >> > From the pdf I found that the minimum distance centroid would be
> the
> >> >> > closest
> >> >> > that I would want. How can I compute it. However it would be better
> >> >> > if
> >> >> > something of the following is provided.
> >> >> > I have a set of points on the surface of earth. I want to calculate
> a
> >> >> > central point such that the distance from it to the given points is
> >> >> > in
> >> >> > required proportion with the additional requirement that this point
> >> >> > should
> >> >> > be inside the polygon obtained by joining the points.
> >> >> > It's easy to see that a such a requirement cannot be perfectly met.
> >> >> > Let
> >> >> > me
> >> >> > give you an analogy in the euclidean geometry rather than spherical
> >> >> > geometry.
> >> >> > Suppose I have 4 points in the Euclidean plane. I want to calculate
> a
> >> >> > point
> >> >> > which is equidistant/distances in equal proportions to the 4
> points.
> >> >> > Now
> >> >> > this is not always possible because if I choose a circle to pass
> >> >> > through
> >> >> > 3
> >> >> > points, it's not necessary that the fourth point too lies on the
> >> >> > circle.
> >> >> > (because 3 points uniquely define a circle.)
> >> >> > So then we can do a kind of multidimensional scaling. Such that we
> >> >> > try
> >> >> > to
> >> >> > find a point which satisfies the criterion as best as it can.
> >> >> > While this is challenging enough to calculate on a euclidean plane.
> >> >> > Spherical geometry makes things even worse!!
> >> >> >
> >> >> > I hope I was able to explain myself. Please do ask me if I was not
> >> >> > clear.
> >> >> >
> >> >> > --
> >> >> > thanks and regards,
> >> >> >
> >> >> > ~Apurv Verma
> >> >> >    B. Tech.(CSE)
> >> >> >    IIT- Ropar
> >> >> >    Mobile - 09256587545
> >> >> >
> >> >> >
> >> >> >
> >> >> >
> >> >> >
> >> >> >
> >> >> > On Mon, Jul 25, 2011 at 1:10 PM, Michael Bedward
> >> >> > <[email protected]>
> >> >> > wrote:
> >> >> >>
> >> >> >> Hello,
> >> >> >>
> >> >> >> Please reply via the list.
> >> >> >>
> >> >> >> So what sort of centroid do you want ? There are many in common
> use
> >> >> >> (see
> >> >> >>
> http://user.gs.rmit.edu.au/rod/files/publications/MSIA_Centroid.pdf
> >> >> >> for a comparison of some).
> >> >> >>
> >> >> >> Or is it enough to have any point that is inside some polygon
> formed
> >> >> >> by the data points e.g. the convex hull ?
> >> >> >>
> >> >> >> Michael
> >> >> >>
> >> >> >>
> >> >> >> On 25 July 2011 16:58, Apurv Verma <[email protected]> wrote:
> >> >> >> > Yes it is a type of centroid calculation only.
> >> >> >> > Given a polygon on the surface of earth, I need to calculate its
> >> >> >> > centroid.
> >> >> >> > Have a look at this page.
> >> >> >> > http://www.fmepedia.com/index.php/InsidePointReplacer
> >> >> >> >
> >> >> >> > thanks and regards,
> >> >> >> >
> >> >> >> > ~Apurv Verma
> >> >> >> >    B. Tech.(CSE)
> >> >> >> >    IIT- Ropar
> >> >> >> >    Mobile - 09256587545
> >> >> >> >
> >> >> >> >
> >> >> >> >
> >> >> >> >
> >> >> >> >
> >> >> >> >
> >> >> >> > On Mon, Jul 25, 2011 at 10:37 AM, Michael Bedward
> >> >> >> > <[email protected]> wrote:
> >> >> >> >>
> >> >> >> >> Hello Apurv,
> >> >> >> >>
> >> >> >> >> >  I am absolutely new to the geo tool library. In fact I want
> to
> >> >> >> >> > use
> >> >> >> >> > it
> >> >> >> >> > for a
> >> >> >> >> > specific purpose in the open source project "phyloGeoRef".
> >> >> >> >>
> >> >> >> >> Great to see taxonomy / biogeography making an appearance here
> :)
> >> >> >> >>
> >> >> >> >> > Here is the
> >> >> >> >> > functionality that I want.
> >> >> >> >>
> >> >> >> >> > Given a set of (lat,long) pairs on the globe. I have to
> >> >> >> >> > calculate
> >> >> >> >> > the
> >> >> >> >> > inside
> >> >> >> >> > point replacer for these set of nodes. Is there some function
> >> >> >> >> > in
> >> >> >> >> > Geo
> >> >> >> >> > Tools
> >> >> >> >> > that provides this functionality. If not, Is there any other
> >> >> >> >> > method
> >> >> >> >> > to
> >> >> >> >> > do it
> >> >> >> >> > ?
> >> >> >> >>
> >> >> >> >> Can you provide more detail about the required output. I'm not
> >> >> >> >> familiar with the term "inside point replacer" and Wikipedia
> >> >> >> >> doesn't
> >> >> >> >> seem to know it either. Is it a type of centroid calculation ?
> >> >> >> >>
> >> >> >> >> Michael
> >> >> >> >
> >> >> >> >
> >> >> >
> >> >> >
> >> >
> >> >
> >
> >
>
------------------------------------------------------------------------------
Got Input?   Slashdot Needs You.
Take our quick survey online.  Come on, we don't ask for help often.
Plus, you'll get a chance to win $100 to spend on ThinkGeek.
http://p.sf.net/sfu/slashdot-survey
_______________________________________________
Geotools-gt2-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users

Reply via email to