[GRASS-user] v.generalize for polygons?

Hamish · February 27, 2008, 2:19am

Hi,

I have a high-res vector area map of regional districts which I wish to generalize. I am having trouble with finding the correct method in v.generalize to use. Currently every thing I try tends to break the area topology and leave only a portion of the now-open boundary.

i.e. it should preserve nodes, only generalize (remove) non-node vertices.
We can assume there is only a single boundary line between areas, so there shouldn't be topological issues. (or at least only in tiny corner cases where the new generalized line overlaps another feature, but that should be easy to fix with v.clean)

the idea is to simplify the map before running v.extrude to get something like this:
http://grass.osgeo.org/grass60/screenshots/images/inc_employ_usa_2002.jpg

but without the 600,000 extruded faces created from every little twist in the coastline.

Hamish

Wolf · February 27, 2008, 10:35am

On 27.02.2008 04:19, Hamish wrote:

Hi,

Hi,

I have a high-res vector area map of regional districts which I wish
to generalize. I am having trouble with finding the correct method in
v.generalize to use. Currently every thing I try tends to break the
area topology and leave only a portion of the now-open boundary.

What methods did you try? What exact commands have you tried that fail? Can you share the problematic map? (you can email it to me directly)

i.e. it should preserve nodes, only generalize (remove) non-node
vertices. We can assume there is only a single boundary line between
areas, so there shouldn't be topological issues. (or at least only in
tiny corner cases where the new generalized line overlaps another
feature, but that should be easy to fix with v.clean)

v.generalize does preserve nodes, and as long as the input map is topologically correct so should the output map be. Perhaps your threshold is way off?

--Wolf

--

<:3 )---- Wolf Bergenheim ----( 8:>

Daniel_Bundala · February 27, 2008, 11:25am

Hi,

I am not sure that I understand correcly what you are trying to do,
but as far as I remember, the boundaries are generalized as polylines.
That is, if the boundary contains some corners, they might be removed
which might create the holes. Also, "ovesimplified" lines are removed
by default. These are the lines that are shorter than the threshold.
However, there is a flag(-r?) which prevents the module from removing
them.

Daniel

On Wed, Feb 27, 2008 at 2:19 AM, Hamish <hamish_b@yahoo.com> wrote:

Hi,

I have a high-res vector area map of regional districts which I wish to generalize. I am having trouble with finding the correct method in v.generalize to use. Currently every thing I try tends to break the area topology and leave only a portion of the now-open boundary.

i.e. it should preserve nodes, only generalize (remove) non-node vertices.
We can assume there is only a single boundary line between areas, so there shouldn't be topological issues. (or at least only in tiny corner cases where the new generalized line overlaps another feature, but that should be easy to fix with v.clean)

the idea is to simplify the map before running v.extrude to get something like this:
http://grass.osgeo.org/grass60/screenshots/images/inc_employ_usa_2002.jpg

but without the 600,000 extruded faces created from every little twist in the coastline.

Hamish
_______________________________________________
grass-user mailing list
grass-user@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/grass-user

Hamish · February 28, 2008, 4:16am

Hamish:

> I have a high-res vector area map of regional districts which I
> wish to generalize. I am having trouble with finding the correct
> method in v.generalize to use. Currently every thing I try tends
> to break the area topology and leave only a portion of the now-
> open boundary.

I have now tried with a related vector, linked below, and it worked
(very!) nicely for that. But it fails with a derivative vector map.
v.digit shows no problems with topography.

Wolf:

What methods did you try?

many of them.. mainly douglas with a number of threshold values.

What exact commands have you tried that fail?

at the simplest: v.generalize in= out=
but some areas are missing.

Can you share the problematic map? (you can email it to me directly)

sure,

starting with:
http://www.stats.govt.nz/statistics-by-area/regional-statistics/geography-mapping/download-digital-boundaries.htm
-- Census based NZMG 2006 (37mb shapefile .zip)

I am looking at regional boundaries (RC) from REGC06_LV2.shp

this map generalizes nicely, but it includes the 12 nautical mile
territorial buffer around the coastline. When I overlay that map with a
detailed coastline is when I see the problem.

I'll send a sample of the v.overlay output off-list.

v.generalize does preserve nodes, and as long as the input map is
topologically correct so should the output map be.

ok. (confirmed, it does a very nice job simplifying the above
shapefile)

Perhaps your threshold is way off?

Possible, as I am just learning. But I did try a number of ranges and
slowly increase. All would be ok for slight generalization then big
breakage.

e.g. it has a big jump between thresh=0.4865 and 0.487

Good:
   v.generalize in=rc_merge_coast3 out=rc_gen thresh=0.4865 --o
   ...
   Number of vertices was reduced from 569815 to 521969 [91%]

Bad:
   v.generalize in=rc_merge_coast3 out=rc_gen thresh=0.487 --o
   ...
   Number of vertices was reduced from 569815 to 336380 [59%]

Daniel:

However, there is a flag(-r?) which prevents the module from removing
them.

Flags:
-c Copy attributes
-r Remove lines and areas smaller than threshold

thanks,
Hamish

____________________________________________________________________________________
Looking for last minute shopping deals?
Find them fast with Yahoo! Search. http://tools.search.yahoo.com/newsearch/category.php?category=shopping

Wolf · March 3, 2008, 3:01pm

Hamish,

I played around with the map you gave me, and I think I found a way around the problem (though I'm still not very sure about what the real problem is, perhaps too many small areas or too many shorter segments (the boundary seems to be built of a number of smaller lines). I'll try to create another problematic map.
Anyway, here is how I was able to generalize it:

# First I fuse the short segments into one long boundary
v.build.polylines --o input=rc_merge_coast3 output=fused
# Then I clean away some (relatively) small islands
v.clean --o in=fused out=clean tool=rmarea thresh=500000.0
# finally generalize
v.generalize -r --o input=clean output=gen method=douglas_reduction reduction=20

The generalized map contains about 20% of the original points.

--Wolf

On 28.02.2008 06:16, Hamish wrote:

Hamish:

I have a high-res vector area map of regional districts which I
wish to generalize. I am having trouble with finding the correct
method in v.generalize to use. Currently every thing I try tends
to break the area topology and leave only a portion of the now-
open boundary.

I have now tried with a related vector, linked below, and it worked
(very!) nicely for that. But it fails with a derivative vector map.
v.digit shows no problems with topography.

Wolf:

What methods did you try?

many of them.. mainly douglas with a number of threshold values.

What exact commands have you tried that fail?

at the simplest: v.generalize in= out=
but some areas are missing.

Can you share the problematic map? (you can email it to me directly)

sure,

starting with:
http://www.stats.govt.nz/statistics-by-area/regional-statistics/geography-mapping/download-digital-boundaries.htm
-- Census based NZMG 2006 (37mb shapefile .zip)

I am looking at regional boundaries (RC) from REGC06_LV2.shp

this map generalizes nicely, but it includes the 12 nautical mile
territorial buffer around the coastline. When I overlay that map with a
detailed coastline is when I see the problem.

I'll send a sample of the v.overlay output off-list.

v.generalize does preserve nodes, and as long as the input map is topologically correct so should the output map be.

ok. (confirmed, it does a very nice job simplifying the above
shapefile)

Perhaps your threshold is way off?

Possible, as I am just learning. But I did try a number of ranges and
slowly increase. All would be ok for slight generalization then big
breakage.

e.g. it has a big jump between thresh=0.4865 and 0.487

Good:
   v.generalize in=rc_merge_coast3 out=rc_gen thresh=0.4865 --o
   ...
   Number of vertices was reduced from 569815 to 521969 [91%]

Bad:
   v.generalize in=rc_merge_coast3 out=rc_gen thresh=0.487 --o
   ...
   Number of vertices was reduced from 569815 to 336380 [59%]

Daniel:

However, there is a flag(-r?) which prevents the module from removing
them.

Flags:
  -c Copy attributes
  -r Remove lines and areas smaller than threshold

thanks,
Hamish

      ____________________________________________________________________________________
Looking for last minute shopping deals? Find them fast with Yahoo! Search. http://tools.search.yahoo.com/newsearch/category.php?category=shopping

--

<:3 )---- Wolf Bergenheim ----( 8:>