[postgis-users] Difficult Problem with Polygons

Nicolas Ribot nicolas.ribot at gmail.com
Mon Oct 29 08:23:47 PDT 2012


Yes, it looks like some points were not put in the right order before
building a polygon, thus these "butterfly" polygons you generated.
If the formed polygon are expected to be convex, you could use
st_convexhull on the point cloud to generate the polygons.

On 29 October 2012 16:09, Ed Linde <edolinde at gmail.com> wrote:
> Thanks Nicolas! Will look at the script that generated the polygon text..
> must have goofed something up there.
>
>
> On Mon, Oct 29, 2012 at 4:05 PM, Nicolas Ribot <nicolas.ribot at gmail.com>
> wrote:
>>
>> Hi,
>>
>> No. this is because some of your polygons are not valid:
>> for instance:
>> select st_isvalid('POLYGON  ((593921 5219610 803,593921 5219610
>> 818,593921 5219620 818,593921 5219620 803,593921 5219610
>> 803))'::geometry);
>> NOTICE:  Too few points in geometry component at or near point 593921
>> 5219610 803
>>  st_isvalid
>> ------------
>>  f
>> (1 row)
>>
>> you can control this with st_isvalid, st_isvalidReason and correct
>> them with st_makeValid.
>>
>> Use only valid objects before processing them with Postgis functions.
>>
>> Nicolas
>>
>> On 29 October 2012 16:03, Ed Linde <edolinde at gmail.com> wrote:
>> > Thanks Nicolas. Just about the error, is this because the line segments
>> > are
>> > too close
>> > and postgis 2.0 could not handle this? If so is there a workaround, even
>> > if
>> > it means
>> > slightly having to "perturb" each point's position to not run into this
>> > bug.
>> > I was really hoping that the intersection of two polygonal shapes in 3D
>> > would be fairly
>> > simple in postgis 2.0 :(
>> >
>> > Ed
>> >
>> >
>> > On Mon, Oct 29, 2012 at 4:00 PM, Nicolas Ribot <nicolas.ribot at gmail.com>
>> > wrote:
>> >>
>> >> If the points are computed in the right order, you can store them and
>> >> pass them to st_makeLine and st_makePolygon.
>> >> If not, you can form a segment between 2 closest points and connect it
>> >> to the closest points.
>> >>
>> >> On 29 October 2012 15:37, Ed Linde <edolinde at gmail.com> wrote:
>> >> > Hi All,
>> >> > Thanks for the tips! Just another thing, when I compute the
>> >> > "transition
>> >> > points" on each edge (shown as red points in my pdf).
>> >> > I need to join them to make a polygon. Wondering how I can connect
>> >> > them
>> >> > together so that I start with a point and end on it
>> >> > to form a closed polygon?
>> >> >
>> >> > Cheers,
>> >> > Ed
>> >> >
>> >> >
>> >> > On Mon, Oct 29, 2012 at 3:03 PM, Stephen Woodbridge
>> >> > <woodbri at swoodbridge.com> wrote:
>> >> >>
>> >> >> Hi Ed,
>> >> >>
>> >> >> Well if clarifies one thing at least, you can ignore Mike's
>> >> >> st_Relate
>> >> >> because 5 is not adjacent to a linear edge, it is only adjacent to a
>> >> >> vertex,
>> >> >> so st_touches should work fine.
>> >> >>
>> >> >> so you are looking for:
>> >> >>
>> >> >> 1. a specific triangle by id
>> >> >> 2. that intersects triangle VC
>> >> >> 3. and the triangles adjacent to triangle by id
>> >> >> 4. and those that are contained in VC
>> >> >> 5. sorted by distance to P1 limit 1
>> >> >>
>> >> >> so something like:
>> >> >>
>> >> >> select id, the_geom
>> >> >>   from (select id, the_geom as adjacent
>> >> >>           from triangles
>> >> >>          where st_touches(the_geom,
>> >> >>                           (select the_geom
>> >> >>                              from triangles a,
>> >> >>                                   (select the_geom as vc
>> >> >>                                      from vc_table
>> >> >>                                     where id='p1') b
>> >> >>                             where a.id=4)) c
>> >> >>  order by st_distance(c.the_geom, (select the_geom
>> >> >>                                      from points
>> >> >>                                     where id='p1')) asc limit 1;
>> >> >>
>> >> >> Untested, but should give you a model to work with.
>> >> >>
>> >> >> -Steve W
>> >> >>
>> >> >>
>> >> >>
>> >> >> On 10/29/2012 6:57 AM, Ed Linde wrote:
>> >> >>>
>> >> >>> Attached is a figure. Where the dotted line is the boundary of the
>> >> >>> voronoi cell whose
>> >> >>> generator is point P1. So triangle "4" intersects with the voronoi
>> >> >>> boundary, but we are
>> >> >>> interested in the adjacent triangles of triangle 4, which are
>> >> >>> closer
>> >> >>> to
>> >> >>> point P1.
>> >> >>> For example, triangle 5.
>> >> >>>
>> >> >>> Hope this helps.
>> >> >>> Cheers,
>> >> >>> Ed
>> >> >>>
>> >> >>> On Mon, Oct 29, 2012 at 11:50 AM, Nicolas Ribot
>> >> >>> <nicolas.ribot at gmail.com
>> >> >>> <mailto:nicolas.ribot at gmail.com>> wrote:
>> >> >>>
>> >> >>>     Could you draw a figure ?
>> >> >>>
>> >> >>>     Nicolas
>> >> >>>
>> >> >>>     On 29 October 2012 11:03, Ed Linde <edolinde at gmail.com
>> >> >>>     <mailto:edolinde at gmail.com>> wrote:
>> >> >>>      > Hi All,
>> >> >>>      > Thanks for the suggestions.
>> >> >>>      > For 1) I will look into how ST_touches works and see if it
>> >> >>> can
>> >> >>>     pick up all
>> >> >>>      > the adjacent polygons to
>> >> >>>      > the one I have. And also look into Mike's suggestion on
>> >> >>>     ST_relate...though I
>> >> >>>      > must admit it looks
>> >> >>>      > more complex.
>> >> >>>      > For 2) I will try to clarify it a bit more... its harder to
>> >> >>> do
>> >> >>>     without a
>> >> >>>      > figure :) but here goes.
>> >> >>>      >
>> >> >>>      > Lets say we have a point Q which is the generator of a
>> >> >>> voronoi
>> >> >>>     cell. Now I
>> >> >>>      > compute the
>> >> >>>      > intersection between the voronoi cell boundaries and my
>> >> >>>     triangulation (Set
>> >> >>>      > of polygons)
>> >> >>>      > using ST_intersect. Once I have these triangles.. I say pick
>> >> >>> one
>> >> >>>     triangle T
>> >> >>>      > that is
>> >> >>>      > intersecting the voronoi cell boundary of Q.
>> >> >>>      > For all the triangles adjacent to T, I need to know which
>> >> >>>     triangles are
>> >> >>>      > INSIDE the voronoi
>> >> >>>      > boundary (closer to Q) and which adjacent triangles are just
>> >> >>>     OUTSIDE the
>> >> >>>      > voronoi
>> >> >>>      > boundary (farther from Q). I am basically testing for a
>> >> >>> certain
>> >> >>>     property by
>> >> >>>      > "shrinking" the
>> >> >>>      > voronoi cell (closer to Q) and another property when
>> >> >>> "expanding"
>> >> >>>     the voronoi
>> >> >>>      > cell (away from Q).
>> >> >>>      > Just need to make this division of triangles. Haven't
>> >> >>> thought
>> >> >>> of
>> >> >>>     a nice way
>> >> >>>      > to do this in postgis 2.0
>> >> >>>      > So any suggestions would greatly help.
>> >> >>>      >
>> >> >>>      > Thanks,
>> >> >>>      > Ed
>> >> >>>      >
>> >> >>>      > On Mon, Oct 29, 2012 at 10:15 AM, Mike Toews
>> >> >>> <mwtoews at gmail.com
>> >> >>>     <mailto:mwtoews at gmail.com>> wrote:
>> >> >>>      >>
>> >> >>>      >> On 29 October 2012 21:33, Ed Linde <edolinde at gmail.com
>> >> >>>     <mailto:edolinde at gmail.com>> wrote:
>> >> >>>      >> > Hi All,
>> >> >>>      >> > I need help with 2 hard problems. I store triangles in a
>> >> >>> table
>> >> >>> as
>> >> >>>      >> > POLYGON.
>> >> >>>      >> >
>> >> >>>      >> > 1. I want to know for a given triangle, which triangles
>> >> >>> share
>> >> >>>     an edge
>> >> >>>      >> > (adjacent) with this triangle.
>> >> >>>      >>
>> >> >>>      >> Sounds like you have a finite element mesh with nodes and
>> >> >>> elements.
>> >> >>>      >> You can use ST_Relate with pattern 'FF2F11212' to pick out
>> >> >>> elements
>> >> >>>      >> that share the same edge. This DE-9-IM is sort-of a custom
>> >> >>>     ST_Touches,
>> >> >>>      >> but only takes linear boundary overlaps. So if you have a
>> >> >>> table
>> >> >>>      >> "elements", and you want to find ones that touch ID 567:
>> >> >>>      >>
>> >> >>>      >> SELECT elements.*
>> >> >>>      >> FROM elements, elements as e
>> >> >>>      >> WHERE e.id <http://e.id> = 567 AND
>> >> >>>
>> >> >>>      >>     ST_Relate(elements.geom, e.geom, 'FF2F11212');
>> >> >>>      >>
>> >> >>>      >> I'm not certain about your second question.
>> >> >>>      >>
>> >> >>>      >> -Mike
>> >> >>>      >> _______________________________________________
>> >> >>>      >> postgis-users mailing list
>> >> >>>      >> postgis-users at postgis.refractions.net
>> >> >>>     <mailto:postgis-users at postgis.refractions.net>
>> >> >>>
>> >> >>>      >>
>> >> >>> http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >> >>>      >
>> >> >>>      >
>> >> >>>      >
>> >> >>>      > _______________________________________________
>> >> >>>      > postgis-users mailing list
>> >> >>>      > postgis-users at postgis.refractions.net
>> >> >>>     <mailto:postgis-users at postgis.refractions.net>
>> >> >>>
>> >> >>>      >
>> >> >>> http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >> >>>      >
>> >> >>>     _______________________________________________
>> >> >>>     postgis-users mailing list
>> >> >>>     postgis-users at postgis.refractions.net
>> >> >>>     <mailto:postgis-users at postgis.refractions.net>
>> >> >>>
>> >> >>>     http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>>
>> >> >>> _______________________________________________
>> >> >>> postgis-users mailing list
>> >> >>> postgis-users at postgis.refractions.net
>> >> >>> http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >> >>>
>> >> >>
>> >> >> _______________________________________________
>> >> >> postgis-users mailing list
>> >> >> postgis-users at postgis.refractions.net
>> >> >> http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >> >
>> >> >
>> >> >
>> >> > _______________________________________________
>> >> > postgis-users mailing list
>> >> > postgis-users at postgis.refractions.net
>> >> > http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >> >
>> >> _______________________________________________
>> >> postgis-users mailing list
>> >> postgis-users at postgis.refractions.net
>> >> http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >
>> >
>> >
>> > _______________________________________________
>> > postgis-users mailing list
>> > postgis-users at postgis.refractions.net
>> > http://postgis.refractions.net/mailman/listinfo/postgis-users
>> >
>> _______________________________________________
>> postgis-users mailing list
>> postgis-users at postgis.refractions.net
>> http://postgis.refractions.net/mailman/listinfo/postgis-users
>
>
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>



More information about the postgis-users mailing list