[postgis-users] Approximate union of polygons?

[Kevin Neufeld]

Have you tried putting your resultant multipolygons through ST_Simplify 
or ST_SimplifyPreserveTopology?  Both use the Douglas-Peuker algorithm 
to weed out unwanted vertices.

Cheers,
Kevin

Dan Johnson wrote:
> I have an application that keeps a database of video footprints, one 
> quadrilateral per frame of video. I consolidate each 5 seconds of 
> video footprints into a single multipolygon in order to speed up 
> queries for video of a specified area. That's typically a union of 150 
> (30fps) polygons with 4-vertexes. I build this multipolygon with 
> ST_GeomUnion, and it's acceptably fast. So far, so good.
>
> I've been quite surprised by the variance in the number of vertices in 
> the resulting multipolygon - sometimes it's as high as 1200. Usually 
> it's lower than the equvalent set of frame footprints, but not always. 
> This hasn't been a problem for database queries, but now it's time to 
> build a user interface and of course the UI has problems displaying 
> that many vertices. (Actually, the UI is getting the polygons via a 
> SOAP message, and our web server's SOAP infrastructure often just dies 
> when the result is so long.) Eyeballing these consolidated polygons, 
> it looks like keeping only 20-30 vertices would provide an excellent 
> approximation. (It's quite common for the camera to stare at a single 
> point for a long time, in which case all the frames are in about the 
> same location.) This seems like the sort of thing PostGIS would be 
> able to do but I haven't been able to figure it out from the docs. 
> Does anyone know if this is possible?
>
> (I've been told the footprints have to consider every frame, so 
> building the union with one frame per second or similar is out.)
>
> Thanks,
> Dan
>
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users