[GRASS-user] Re: grass-user Digest, Vol 47, Issue 52
Micha Silver
micha at arava.co.il
Thu Apr 1 08:23:12 EDT 2010
On 04/01/2010 02:52 PM, Nathaniel Iwuchukwu wrote:
> i created 11 polygons from different portions of the road.
So if I understand correctly, the procedure is working, but you're not
getting the results you'd expect?
Maybe you could supply to the grass-users list some URL's of images of
what the satellite rasters look like, with your training polygons, and
the resulting roads. Then perhaps someone will be able to suggest how to
get more complete results.
>
>
> 2010/4/1 Micha Silver <micha at arava.co.il <mailto:micha at arava.co.il>>
>
> On 04/01/2010 01:11 AM, Nathaniel Iwuchukwu wrote:
>> Thanks Micha,
>> I ran r.thin and got just one line. is there a way i can improve it?
>>
> Good, so we're making progress...
> How many polygon areas did you create in the training map? Does it
> cover various portions of the roads?
>> On Wed, Mar 31, 2010 at 1:54 PM, Micha Silver <micha at arava.co.il
>> <mailto:micha at arava.co.il>> wrote:
>>
>> Nathaniel Iwuchukwu wrote:
>>
>> Also, when i convert from vector to raster, i don't see
>> anything on the new raster map. its all blank. what
>> should i do. thanks
>>
>> I guess that would explain why you're getting "Training map
>> has no classes".
>> Can you display the vector which covers sections of the road?
>> Do you see filled polygons?
>> If not maybe you made boundaries with no centroids? In GRASS,
>> area features must be composed of both a boundary *and* a
>> centroid. In the digitizer there's a special button for
>> adding a centroid. You need one inside each polygon boundary.
>>
>>
>> On Wed, Mar 31, 2010 at 11:20 AM, Nathaniel Iwuchukwu
>> <ikechinedu at gmail.com <mailto:ikechinedu at gmail.com>
>> <mailto:ikechinedu at gmail.com
>> <mailto:ikechinedu at gmail.com>>> wrote:
>>
>> Thanks Micha,
>> I will definately post it once i scale through it.
>> i am getting an error when i run the i.gensigset on
>> the training data.
>> this is what i typed and the result:
>> GRASS 6.4.0RC5 (sis):~/Desktop > i.gensigset
>> trainingmap=traingis
>> group=sislabgp subgroup=sislabsub signaturefile=sislab.sig
>> Finding training classes...
>> 100%
>> ERROR: Training map has no classes
>> do you have any clue why i don't have a class?
>> Thanks Man
>> 2010/3/31 Micha Silver <micha at arava.co.il
>> <mailto:micha at arava.co.il> <mailto:micha at arava.co.il
>> <mailto:micha at arava.co.il>>>
>>
>>
>> On 03/31/2010 01:19 AM, Nathaniel Iwuchukwu wrote:
>>
>> Thanks Micha,
>> this looks good but am still working on it.
>> would you know
>> the RGB for a road? from the link you
>>
>> No, I can't tell what the RGB values for a road
>> would be, and
>> it will change from image to image, and from road
>> to road.
>>
>> sent me, i will need it to capture the pixel
>> values of road
>> polygons that i am digitizing.
>>
>> You *don't* need to find the actual pixel values
>> for the road
>> polygons. You just digitize strips that cover the
>> road, and
>> feed these areas (after converting to a raster) as the
>> training map. The gensigset module creates the
>> collection of
>> pixels values that match "roads"
>>
>> thanks Micha. you really gave me a lead.
>>
>> Good. Post back to the list when you get it worked
>> out.
>>
>>
>> Nathaniel
>>
>> On Tue, Mar 30, 2010 at 4:26 PM, Micha Silver
>> <micha at arava.co.il <mailto:micha at arava.co.il>
>> <mailto:micha at arava.co.il
>> <mailto:micha at arava.co.il>>> wrote:
>>
>> Nathaniel Iwuchukwu wrote:
>>
>> sorry my typo. i mean classification.
>>
>> I didn't intend to insult. Sometimes typos
>> come out
>> funny, and I can't resist pointing it out...
>>
>> but am looking for procedure on road
>> extraction from
>> satellite imagery with high resolution
>>
>> OK, you'll probably want to use "supervised"
>> classification. This means giving the
>> classification
>> algorithm a "training" layer. So the
>> process involves
>> 1- First create a vector layer using the
>> grass digitizer
>> with small polygons covering some sections
>> of the roads
>> in the satellite image. Convert the vector
>> to a raster
>> (v.to.rast). This will be the training map.
>> 2- Now group the satellite bands into an
>> imagery group
>> (i.group)
>> 3- Run i.gensigset with the training
>> raster and the
>> imagery group.
>> 4- Now use i.smap with the signature file
>> and imagery
>> group as input, and a classification map
>> as output.
>> 5- If the results look good, run r.thin on
>> this raster,
>> and then r.to.vect to see what you get.
>>
>> Here's a more detailed explanation that
>> you might find
>> helpful:
>> http://www.surfaces.co.il/?p=3
>>
>> Regards,
>> Micha
>>
>>
>>
>> -- Micha Silver
>> http://www.surfaces.co.il/
>> Arava Development Co. +972-52-3665918
>>
>>
>>
>>
>> This mail was received via Mail-SeCure System.
>>
>>
>>
>> --
>> Micha Silver
>>
>> Arava Development Co. +972-52-3665918
>> http://www.surfaces.co.il <http://www.surfaces.co.il/>
>>
>>
>>
>>
>> This mail was received via Mail-SeCure System.
>
>
> --
> Micha Silver
> http://www.surfaces.co.il/
> Arava Development Co. +972-52-3665918
>
>
>
> This mail was received via Mail-SeCure System.
--
Micha Silver
http://www.surfaces.co.il/
Arava Development Co. +972-52-3665918
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.osgeo.org/pipermail/grass-user/attachments/20100401/6999acf0/attachment.html
More information about the grass-user
mailing list