[postgis-users] RE: WKT Raster project: call for fundings

Fri Dec 5 13:49:42 PST 2008

Sandro, Pierre:

I am part of an academic research institute that has need of the WKT
Raster functionality described by Pierre's email and presentation.

Our use case is as follows:
We are developing a web service for disseminating estimates of wildland
fire emissions.  The primary data inputs are fire perimeters (vector)
and a map of the fuels (raster). The raster is a 1km grid covering North
America, with 10+ bands that characterize the type of fuel available
(i.e. duff, understory, midstory, nonwoody, etc.).  Basically we would
like to intersect the fire perimeter vectors with the fuel raster, and
then pass the results to an fire emissions model.

Our preliminary design is to use postgis for data storage and django
(python web framework) for gluing together the processing components. 
And to get around the current lack of vector+raster spatial operations,
we will start by converting the fuel raster to a large number of point
geometry records (7,741,499 to be exact) so that we can store it in
postGIS and utilize the intersection function for points and polygons. 
It's certainly not the most elegant design, and one that won't scale if
we move to a temporally changing fuel layer (updated by remote sensing
data).

I think our use case falls well outside the 'stupid images-in-database'
group designation.  At present, we don't plan to display the fuel data
at all, but rather will use the data to feed the emissions model.

We have other similar use cases.  Our research institute is primarily
focused on remote sensing technologies, so we often have a need to
integrate multi-band images (rasters) with reference data or classified
objects (vectors).

So in short, we are interested in seeing the WKT Raster functionality
move forward, and would be interested in co-funding this work.  You can
contact me directly to continue the discussion.

- Tyler

-- 
Tyler A. Erickson, Ph.D.
Research Scientist, Michigan Tech Research Institute, and 
Adjunct Assistant Professor, Dept. of Civil & Environmental Engineering
Michigan Technological University
3600 Green Court, Suite 100
Ann Arbor, MI 48105
W 83.6889°, N 42.3021° (WGS84) 
tyler.erickson at mtu.edu 
(734) 913-6846
http://people.mtri.org/tyler+erickson
http://www.mtri.org
http://www.michiganview.org

Pierre Racine wrote:
> Hi all,
>
> For myself, I will start to work on the project about 15 hours per week in march 2009. At this rhythm, we won't have a beta before winter 2010. This is why help would be welcome.
>
> So two programmers would be working on the project if someone can fund Sandro. One full time (Sandro), one half time (me). For now Sandro has very few money and I'll have few time.
>
> The main idea of WKT Raster is to treat raster tiles as much as possible like PostGIS geometries so we could query and analyse (intersection, overlay, interpolation, conversion, etc...) vectors and rasters seamlessly. WKT Raster is much simpler than Oracle Georaster and much more GIS oriented (like PostGIS is).
>
> A PowerPoint of the design is available at http://www.cef-cfr.ca/uploads/Membres/WKTRasterPostGIS.ppt
>
> Pierre
>
>   
>> -----Message d'origine-----
>> De : Sandro Santilli [mailto:strk at keybit.net] 
>> Envoyé : 4 décembre 2008 06:21
>> À : postgis-devel at postgis.refractions.net; 
>> postgis-users at postgis.refractions.net
>> Cc : Pierre Racine
>> Objet : WKT Raster project: call for fundings
>>
>> Hi all,
>>
>> As some of you may know, I've been working the past two years
>> on the GNU Flash Player [1], something I've felt the need for
>> since I've been doing flash maps for a few years w/out a fully
>> free stack supporting the result (playback).
>>
>> Now Gnash is good enough for use when authors are willing to 
>> target it, and sponsorship reduced, so I'm back
>> lurking in the GIS field.
>>
>> Doing so, I've found the new WKT-raster thread started
>> by Pierre Racine [2] and tought that's something I might be
>> available to work on, given enough fundings can be found.
>>
>> The first steps in that direction would be:
>>
>> - Define a new RASTER postgresql type.
>> - Implement canonical input and output SQL routines.
>> - Implement an importer producing a table
>>   of RASTER tiles from a single raster file.
>> - Implement an SQL aggregate to return a jpeg from a set.
>> - Overlap operator and simple inspectors (bounds, pixels,
>>   bands, sizes).
>>
>> I'm working with Pierre to further refine those steps,
>> but you get the idea. It would take about 6 months
>> to complete the whole set, for an ideal target of 30K EUR.
>>
>>  So, who's interested in (co)funding the work ?
>>
>> I've a partial contribution, but would need 10x more to go.
>> Let's see if the PostGIS community can do it again [3]
>> ( and better this time :)
>>
>> [1] http://www.gnu.org/software/gnash
>> [2] 
>> http://postgis.refractions.net/pipermail/postgis-users/2008-Jul
>> y/020486.html
>> [3] http://blog.cleverelephant.ca/2005/10/concurrency-for-postgis.html
>>
>> --strk;
>>
>>
>> ()   ASCII Ribbon Campaign
>> /\   Keep it simple! 
>>
>>
>>     
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>
>   

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20081205/1cc0d49e/attachment.html>