[postgis-devel] Stats

Pierre Racine Pierre.Racine at sbf.ulaval.ca
Mon Dec 3 08:10:16 PST 2012


Can't you just consider the extra bands the z or t dimensions? Then everything is just a matter of display or interpretation (as a voxel for example). 

Problems are: 

1) we don't have a way to tile in this extra dimension (i.e. our georeferencing system is 2d only, you could put the base altitude for a tile (and why not time) in an extra column though), 

2) we are limited to uint16 bands so a maximum of 65536 values in the z or t dimensions per tile.

3) we don't have any function doing something useful on such a haphazard structure.

So I guess if you add an altitude and a time column to a raster table, you can store 4-d data. Band numbers have to be considered meters offset from base altitude, unless you also store the pixel size in the z dimension in an extra column. Then you can build and API taking these extra columns into accounts to report something useful.

But still that would be a 3-d (or a 4-d) system. Not a n-d.

We need a multi-dimensional array type. Like super Rasdaman "The ultimate Raster Server" (http://www.rasdaman.com/).

But I would think a 4 dimensions array (x, y, z, t) would be sufficient for most geospatial applications. The question is how many regularly gridded dimensions do you need to be stored as a user defined type? I think even just 3 is necessary. Time is probably not something you want to grid.

Pierre

> -----Original Message-----
> From: postgis-devel-bounces at lists.osgeo.org [mailto:postgis-devel-
> bounces at lists.osgeo.org] On Behalf Of Paul Ramsey
> Sent: Saturday, December 01, 2012 3:36 PM
> To: PostGIS Development Discussion
> Subject: Re: [postgis-devel] Stats
> 
> No, unfortunately it doesn't work like that. We need a raster system
> that can do voxels, not just pixels. If you're thinking about your
> image coordinates in terms of i,j, or x,y, you aren't going to be able
> to do N-D, no matter how many bands you have available. (Well, not
> entirely true, you could hack a 3-d system by having band one be the z
> coordinate and band two be the z value, but that's as far as that
> trick will take you.)
> 
> P.
> 
> On Sat, Dec 1, 2012 at 7:51 AM, Sandro Santilli <strk at keybit.net> wrote:
> > On Fri, Nov 30, 2012 at 05:19:08PM -0800, dustymugs wrote:
> >> I guess my initial answer would be: No, not in the same context as in
> >> geometry.
> >>
> >> Having said that, things like elevation datasets (NED) could have the
> >> band pixel values considered the Z axis while the X and Y axes are the grid.
> >>
> >> I don't know what the fourth dimension would be... time?
> >
> > How about a dimension each band ?
> > Would be a minimum of 3 dimensions per raster.
> >
> > Can we have that Paul ?
> >
> > --strk;
> > _______________________________________________
> > postgis-devel mailing list
> > postgis-devel at lists.osgeo.org
> > http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-devel
> _______________________________________________
> postgis-devel mailing list
> postgis-devel at lists.osgeo.org
> http://lists.osgeo.org/cgi-bin/mailman/listinfo/postgis-devel



More information about the postgis-devel mailing list