[GRASS-user] i.segment and i.segment.stats errors

[Jamille Haarloo]

first lines of test.csv

cat|area|perimeter|compact_square|compact_circle|fd
1|11876.000000|1398.000000|0.311808|3.618821|1.543932
2|635805.000000|57572.000000|0.055400|20.367814|1.640512
3|25898.000000|3322.000000|0.193773|5.823198|1.595825
4|7767.000000|1702.000000|0.207122|5.447885|1.661054
5|13605.000000|2200.000000|0.212074|5.320696|1.617158
6|442366.000000|44230.000000|0.060150|18.759495|1.645730

--
Windows 7
Dell
Proc. Intel(R) Core(TM) i7-6700 CPU @ 3.40GHz
RAM: 16 GB
--
Could the "ju" be caused my the Mask?
--
I used uspo on 3 regions to decide on parameters. 0.24 as treshold and
minsize 3500 looked to be the best combination of parameters studying all
the results in the output file.
Summary:
regions=regionAB at LUP1,regionD at LUP1,regionRM at LUP1
segmentation_method=region_growing threshold_start=0.18 threshold_stop=0.25
threshold_step=0.02 minsize_start=3000 minsize_stop=6000 minsize_step=500
optimization_function=f number_best=2 memory=8000 processes=4
Working on region regionAB at LUP1
Working on region regionD at LUP1
Working on region regionRM at LUP1
Best values:
Region Thresh Minsize Optimization
regionRM at LUP1 0.24 3500 0.7951456323
regionRM at LUP1 0.24 3000 0.765849594003
regionAB at LUP1 0.24 4000 0.630706501774
regionAB at LUP1 0.24 3500 0.472619083853
regionD at LUP1 0.2 3000 0.802618475115
regionD at LUP1 0.22 3000 0.77939730321

I also had other uspo runs but with less useful results.

The map with segments was created (0.24 treshold minsize 3500). But after
the errors with i.segment.stats I reasoned the number of segments (18685)
was too much to handle as it failed to create the attribute table. To
reduce the number of segments, I used this map as seed to create a new
segmentation map with the parameters 0.25 and 5000 and a mask to eliminate
the matrix.
But now I see the number of segments shouldn't have been a problem in this
case.

On Tue, Oct 9, 2018 at 9:35 AM Moritz Lennert <mlennert at club.worldonline.be>
wrote:

> On 09/10/18 13:48, Jamille Haarloo wrote:
> > Hello Moritz,
> >
> > First a pop-up came about not finding it, so reinstalled it.
> >
> > g.extension r.object.geometry
> > WARNING: Extension <r.object.geometry> already installed.
> Re-installing...
> > Downloading precompiled GRASS Addons <r.object.geometry>...
> > Updating addons metadata file...
> > Installation of <r.object.geometry> successfully finished
> > (Mon Oct 08 16:55:04 2018) Command finished (4 sec)
> > (Mon Oct 08 16:55:53 2018)
> > r.object.geometry SegmW24IDM4DV4 output=test.csv
> > Calculating statistics
> > Writing output
> > (Mon Oct 08 16:56:14 2018) Command finished (21 sec)
>
> Could you provide the first lines of content of test.csv ?
>
> >
> > g.region -p
> > projection: 1 (UTM)
> > zone:       21
> > datum:      wgs84
> > ellipsoid:  wgs84
> > north:      634826.41313284
> > south:      611477.61329757
> > west:       775494.11663193
> > east:       793630.22784822
> > nsres:      0.50000642
> > ewres:      0.50000307
> > rows:       46697
> > cols:       36272
> > cells:      1693793584
>
>
> So you have over 1.5 billion cells. That's quite big, but not impossible
> to handle. Depends on your machine.
>
> >
> > r.info <http://r.info> SegmW24IDM4DV4
> >
>  +----------------------------------------------------------------------------+
> >   | Map:      SegmW24IDM4DV4                 Date: Fri Oct 05 23:46:04
> > 2018    |
> >   | Mapset:   LUP1                           Login of Creator: haarlooj
> >        |
> >   | Location: Cottica_LUP
> >        |
> >   | DataBase: Z:\GRASS_GIS
> >         |
> >   | Title:
> >         |
> >   | Timestamp: none
> >        |
> >
>  |----------------------------------------------------------------------------|
> >   |
> >        |
> >   |   Type of Map:  raster               Number of Categories: 0
> >         |
> >   |   Data Type:    CELL
> >         |
> >   |   Rows:         46697
> >        |
> >   |   Columns:      36272
> >        |
> >   |   Total Cells:  ju
>
>
> This 'ju' is weird, here.
>
>
> >         |
> >   |        Projection: UTM (zone 21)
> >         |
> >   |            N: 634826.41313284    S: 611477.61329757   Res:
> > 0.50000642      |
> >   |            E: 793630.22784822    W: 775494.11663193   Res:
> > 0.50000307      |
> >   |   Range of data:    min = 1  max = 12901
> >         |
> >   |
> >        |
> >   |   Data Description:
> >        |
> >   |    generated by i.segment
> >        |
> >   |
> >        |
> >   |   Comments:
> >        |
> >   |    i.segment --overwrite group="DV4 at LUP1,IDM4 at LUP1,W2 at LUP1,W4 at LUP1"
> > out\   |
> >   |    put="SegmW24IDM4DV4" threshold=0.25 radius=1.5
> > method="region_growin\   |
> >   |    g" similarity="euclidean" minsize=5000 memory=10000
> > seeds="SegW24IDM\   |
> >   |    4DV4 at LUP1" goodness="SegmW24IDM4DV4_goodn"
>
>
> You only have 12901 segments, so an average size of 32822m2 per segment.
> Is this what you aimed for ? As you asked for a minsize of 5000 pixels I
> guess this is deliberate.
>
> You threshold of 0.25 is also very high...
>
> Moritz
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/grass-user/attachments/20181009/006035d9/attachment.html>