[postgis-users] Raster pixel value

[Andreas Forø Tollefsen]

Update:

I think my suspicion is correct. If I do a ST_Summarystats().sum and divide
this on 36 my MAX value will be 1.
Hence, I think the number of values counted and the number of observations
counted is not equal.

New query:
DROP TABLE IF EXISTS mountain_phil_cell;

SELECT
a.gid As gid,
(ST_SummaryStats((ST_Union(ST_MapAlgebraExpr(ST_AsRaster(a.cell, b.rast,
'32BF'), b.rast, 'rast2', '32BF','INTERSECTION','0','0',0))).rast,
false)).sum / 36 As avgmnt
INTO mountain_phil_cell
FROM
priogrid_land a LEFT JOIN
mountain_phil b
    ON ST_Intersects(a.cell, b.rast)
GROUP BY a.gid
ORDER BY a.gid;


2011/11/25 Andreas Forø Tollefsen <andreasft at gmail.com>

> A small note regarding this issue.
>
> My problem is that I never get a mean value of 1 even if all pixels inside
> the geometry is one.
>
> Could this be because: 6x6 pixels goes into one polygon when visually
> controlling. If each pixel has the value 1, then this will be calculated as
> 36 / 36 = 1. However, if it calculates the sum to be 36 and divide by a
> number higher than 36 pixels, then the result will always be below 1.
> What i am thinking is that while it sums up the pixel values correctly, it
> does not count only the 36 pixels, but also neighboring pixels. Therefore:
> 1+1+1+1...n36 / Number of pixels higher than 36 will always lead to a
> number lower than 1.
>
> Anyone who knows the functions well could probably answer this.
>
> Best regards,
> Andreas
>
>
> 2011/11/25 Andreas Forø Tollefsen <andreasft at gmail.com>
>
>> Could this have to do with the tiling of the raster?
>> I will try to run the same query with a untiled mountain raster to see if
>> that changes anything.
>>
>> Btw. When loading a tiled postgis raster into qgis it shows up with many
>> artifacts and no data areas. The same raster untiled does not show up the
>> same way.
>> Qgis bug?
>>
>> Andreas
>>
>> 2011/11/25 Andreas Forø Tollefsen <andreasft at gmail.com>
>>
>>> Hi,
>>>
>>> Thanks for all of the suggestions. I will do some more testing. However,
>>> as for suggestion 1 i think the pixel size should be the same as the
>>> original raster or am I wrong?
>>>
>>> Both the mean_mnt_bin raster and the priogrid_land shapefile can be
>>> downloaded as zip (2 mb) here:
>>> http://gisintersect.com/mean_mnt_bin.zip
>>> http://gisintersect.com/priogrid_land.zip
>>>
>>> Any help on getting the correct values would be very much appreciated.
>>>
>>> My query:
>>> DROP TABLE IF EXISTS mountain_cell;
>>>
>>> SELECT
>>> a.gid As id,
>>> (ST_SummaryStats((ST_Union(ST_MapAlgebraExpr(ST_AsRaster(a.cell, b.rast,
>>> '32BF'), b.rast, 'rast2', '32BF','INTERSECTION','0','0',0))).rast,
>>> false)).mean As avgmnt
>>> INTO mountain_cell
>>> FROM
>>> priogrid_land a LEFT JOIN
>>> mountain b
>>>     ON ST_Intersects(a.cell, b.rast)
>>> GROUP BY a.gid
>>> ORDER BY a.gid;
>>>
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/postgis-users/attachments/20111125/4efb7d70/attachment.html>