[postgis-users] point to raster conversion

[Rémi Cura]

Is there a special trick to display the raster in QGIS?
I do
db manage/right click on the raster table/add to canvas
Seems like there is nothing in it afterward.

Thanks again,

Rémi-C


2013/12/19 Rémi Cura <remi.cura at gmail.com>

> Buffering make it works.
>
> Cheers,
> Rémi-C
>
>
> 2013/12/19 Rémi Cura <remi.cura at gmail.com>
>
>> Hey, thanks for the answer, and nice catch !
>> I'll try to simply buffer points,
>> then try th eother method to count
>>
>> About performance :
>> this is not about plpgsql or C.
>> I did quit the same processing when importing files of 3 millions points
>> to split into small 1*1*1 m cubes.
>> There were a lot of cubes (around 2k for 3 millions points), a lot of
>> points (3 millions) with a lot of attributes (around 20 flaot/double per
>> point), and it was around 60 sec/ file (for the data processing. Data
>> reading from disk was little longer).
>>
>>
>> For a 50x50 pixels and a few thousand points I would expect around 200ms.
>>
>>
>> I suspect it all boils down to current pixel access mechanism.
>> *Can you please confirm if there is or isn't a method to set/get several
>> pixels at a time?*
>>
>> Thanks anyway,
>> I wouldn't have found =)
>>
>> Cheers,
>> Rémi-C
>>
>>
>> 2013/12/19 Pierre Racine <Pierre.Racine at sbf.ulaval.ca>
>>
>>> "MEAN_OF_VALUES_AT_PIXEL_CENTROID", which is the default
>>> ST_ExtractToRaster() method, search for geometries intersecting with the
>>> centroid of the pixel (a point). That's the method you use both calls to
>>> ExtractToRaster(). As it's almost impossible that a point intersects a
>>> point, this is certainly not the right method to get the value of points
>>> inside the pixel. A proper method would be "MEAN_VALUE_OF_POINTS" or
>>> "MEAN_VALUE_OF_GEOMETRIES" or "FIRST_GEOMETRY_VALUE" which are not
>>> implemented...
>>>
>>> I suggest you try 'COUNT_OF_POINTS' for now which will set each pixel
>>> value to the number of point intersecting with it. If you get it to work
>>> and you're satisfied with the performance I could help adding a method for
>>> your specific need.
>>>
>>> Can't do anything about the performance. All this is pure pl/pgsql and
>>> it's as fast as it can. 5 seconds to compute 2500 pixels value using SQL is
>>> not that bad though if you're not on the web... Otherwise you should have a
>>> specific C function doing exactly what you want. Expect time and $$$.
>>>
>>> Pierre
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20131219/ba0dd2cc/attachment.html>