[postgis-users] raster regions

[Joao Cordeiro]

Em 18/07/2019 10:12, Pierre Racine escreveu:
>
> What is your goal : to simplify the queries expression or to optimize 
> the processing time?
>

The goal is a bit theoretic, focused on gainning expressiveness for 
query languages, by joinning both query and a map algebra language 
together. so that regions in the raster side could be identified by 
algebraic expressions. Some topological assumptions must be considered 
so the nine-interceptions method can be applied with no format 
conversion. I guess PostGIS can be a platform to go around the issue.

> Yes it would work with touch as well.
>
> At some point you have to convert pixels to something that is 
> comparable with vector. The point is to lower this conversion to the 
> minimum. The trick in PostGIS it to tile everything so that only 
> required tiles are converted. You could certainly have some better 
> algorithms than what ST_Intersects(d.geom, e.rast) is doing to process 
> at the pixel level instead of at the tile level.
>
> Pierre
>
> *De :*Joao Cordeiro <joao.cordeiro at inpe.br>
> *Envoyé :* 17 juillet 2019 10:30
> *À :* Pierre Racine <Pierre.Racine at sbf.ulaval.ca>
> *Objet :* Re: [postgis-users] raster regions
>
> Hi, Pierre,
>
> Thanks, I appreciate your replies .
>
> That is quite a clear and readable solution to describing an 
> i*ntercepting* situation. Should it work with *st_touch *as well? 
> Without any raster-vector conversion for the data? My "consume dream" 
> is to have it all in a single query expression, in which the SQL 
> parser delegates to a map algebra parser the evaluation of a kind of 
> "pertinence criteria" to regions described through "map algebraic" 
> expressions. I guess, usinga binary representation to "identify" the 
> regions will allways break an user question into two! For instance, a 
> question such as "an specific natural event (such as flooding) 
> touching a given municipality", would allways demmand for some 
> (tricky) adequate binary data, prior to being expressed in the query 
> language.
>
> Regards
>
> Em 16/07/2019 10:05, Pierre Racine escreveu:
>
>     Right. To minimize the amount of intermediate computation you can
>     add a discriminant to the first query:
>
>     WITH floodareas AS (
>
>       SELECT ST_MapAlgebra(rast, ‘1BB’, ‘CASE WHEN [rast] < 10 THEN 1
>     ELSE NULL END’) rast
>
>       FROM elevation e, districts d
>
>       WHERE ST_Intersects(d.geom, e.rast)
>
>     )
>
>     SELECT d.*
>
>     FROM districts d, floodareas f
>
>     WHERE ST_Intersects(d.geom, f.rast)
>
>     Pierre
>
>     *De :*postgis-users <postgis-users-bounces at lists.osgeo.org>
>     <mailto:postgis-users-bounces at lists.osgeo.org> *De la part de*
>     joao.cordeiro at inpe.br <mailto:joao.cordeiro at inpe.br>
>     *Envoyé :* 15 juillet 2019 14:40
>     *À :* postgis-users at lists.osgeo.org
>     <mailto:postgis-users at lists.osgeo.org>
>     *Objet :* Re: [postgis-users] raster regions
>
>     Thanks, Pierre,
>
>     It seems ok to model the query the way you showed, however, that
>     implies the generation of an entire intermediate binary raster,
>     just to identify the region inside the whole mapped area. My issue
>     is on avoiding such intermediate data by incorporating map algebra
>     expression into the same query expression, Another example, ""...
>     d.geom touch (ndvi2018 < ndvi2019)" where ndvixxxx are raster data.
>
>     Att.
>
>     Em 15-07-2019 14:43, Pierre Racine escreveu:
>
>         Hi,
>
>         That could translate into something like this in PostGIS:
>
>         WITH floodareas AS (
>
>         SELECT ST_MapAlgebra(rast, ‘1BB’, ‘CASE WHEN [rast] < 10 THEN
>         1 ELSE NULL END’) rast
>
>         FROM elevation
>
>         )
>
>         SELECT d.*
>
>         FROM districts d, floodareas f
>
>         WHERE ST_Intersects(g.geom, f.rast)
>
>         NULL becomes nodata values. You could also use ST_Reclass()
>         instead of ST_Mapalgebra().
>
>         Pierre
>
>         *De :*postgis-users <postgis-users-bounces at lists.osgeo.org
>         <mailto:postgis-users-bounces at lists.osgeo.org>> *De la part
>         de* Joao Cordeiro
>         *Envoyé :* 15 juillet 2019 10:19
>         *À :* 'PostGIS Users Discussion'
>         <postgis-users at lists.osgeo.org
>         <mailto:postgis-users at lists.osgeo.org>>
>         *Objet :* [postgis-users] raster regions
>
>         Hi there,
>
>         I am not actually a POSTGIS user, or developer, although I am
>         doing some research of integration of vector -raster in a dbms
>         context. My main question s about identifying regions
>         represented in a domain like the "hybrid raster" so that they
>         can be referred to in a query. For example, I would like to
>         write something like "/*select d from districts where
>         d.geometry touches flooding*/", where "flooding" indicates an
>         expression intended to decide on the pertinence of locations
>         to a flooding ​​risk area described, for instance through map
>         algebra.
>
>         Thanks for any hints or suggestions.
>
>         Att.
>
>         _______________________________________________
>
>         postgis-users mailing list
>
>         postgis-users at lists.osgeo.org  <mailto:postgis-users at lists.osgeo.org>
>
>         https://lists.osgeo.org/mailman/listinfo/postgis-users
>
>
>
>     _______________________________________________
>
>     postgis-users mailing list
>
>     postgis-users at lists.osgeo.org  <mailto:postgis-users at lists.osgeo.org>
>
>     https://lists.osgeo.org/mailman/listinfo/postgis-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20190719/fdc54726/attachment.html>