Ideas on qualifying urban shapes: linear / circular / star

classic Classic list List threaded Threaded
11 messages Options
Reply | Threaded
Open this post in threaded view
|

Ideas on qualifying urban shapes: linear / circular / star

Mathieu Rajerison
Hello list,


I have determined major urban areas.

This is just a post to get ideas from R-users on how to qualify urban
shapes.

The data can either be binary raster (urban/ not urban), either vector.

1) Some urban areas follow linear infrastructures, thus are linear
2) Some other diverge equally from a central heart, and are circular.
3) Some are a mix and are like stars.

The idea would be to get an index that give for each area, the probability
of belonging to each of these three classes. Like a 3-column data frame

I wondered if packages already existed, or statistical methods for this
purpose. Notably, I think that topographic derivatives derived from
smoothed/unsmoothed binary data like aspect, could be used to qualify these
shapes (?)

Thanks for any idea or exchange on the subject!


Mathieu

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Barry Rowlingson
On Fri, Sep 30, 2011 at 10:21 AM, Mathieu Rajerison
<[hidden email]> wrote:

> Hello list,
>
>
> I have determined major urban areas.
>
> This is just a post to get ideas from R-users on how to qualify urban
> shapes.
>
> The data can either be binary raster (urban/ not urban), either vector.
>
> 1) Some urban areas follow linear infrastructures, thus are linear
> 2) Some other diverge equally from a central heart, and are circular.
> 3) Some are a mix and are like stars.
>
> The idea would be to get an index that give for each area, the probability
> of belonging to each of these three classes. Like a 3-column data frame
>
> I wondered if packages already existed, or statistical methods for this
> purpose. Notably, I think that topographic derivatives derived from
> smoothed/unsmoothed binary data like aspect, could be used to qualify these
> shapes (?)
>
> Thanks for any idea or exchange on the subject!

 Sounds like shape analysis or morphometrics. There's package:shape
that might help. Sometimes these techniques depend on landmarks though
(like the locations of eyes, noses, and ears on faces) and probably
won't be any use on vector or raster datasets.

http://en.wikipedia.org/wiki/Morphometrics

Barry

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Sarah Goslee
In reply to this post by Mathieu Rajerison
Hi,

On Fri, Sep 30, 2011 at 5:21 AM, Mathieu Rajerison
<[hidden email]> wrote:

> Hello list,
>
>
> I have determined major urban areas.
>
> This is just a post to get ideas from R-users on how to qualify urban
> shapes.
>
> The data can either be binary raster (urban/ not urban), either vector.
>
> 1) Some urban areas follow linear infrastructures, thus are linear
> 2) Some other diverge equally from a central heart, and are circular.
> 3) Some are a mix and are like stars.

That sounds like the kind of task that patch metrics such as perimeter/area
ratio and fractal dimension were created for. Take a look at the copious
Fragstats literature. I don't know if any have been implemented in R, but
wouldn't be surprised.

Sarah

> The idea would be to get an index that give for each area, the probability
> of belonging to each of these three classes. Like a 3-column data frame
>
> I wondered if packages already existed, or statistical methods for this
> purpose. Notably, I think that topographic derivatives derived from
> smoothed/unsmoothed binary data like aspect, could be used to qualify these
> shapes (?)
>
> Thanks for any idea or exchange on the subject!
>


--
Sarah Goslee
http://www.functionaldiversity.org

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Mathieu Rajerison
Hi,


Maybe you're right: perimeter/ratio could be sufficient...

Fractal dimension and lacunarity are good indicators for urban areas. Do you
know any R package, tool to quantify these?

I have found an ImageJ plugin called Fraclac.

There is another one called r.lacunarity included in spatialtools
http://www.clusterville.org/spatialtools/index.html
r.lacunarity is interesting compared to ImageJ::Fraclac because it uses a
moving window and generates a raster.

I launched a post about lacunarity and fractal dimension on R but didn't
have any answer.


So, if anyone manages to use r.lacunarity or knows other tools than Fraclac,
I'd be happy!


For those interested , here is some literature on fractal dimension applied
to analysis or aerial images or cities:
http://www.public.asu.edu/~smyint/publications/CEUS-Lacunarity-Myint-Lam.pdf
http://cybergeo.revues.org/8902
http://www.isprs.org/proceedings/XXXVII/congress/3b_pdf/80.pdf
http://www.fatih.edu.tr/~mcadams/geo352/fractal.pdf


2011/9/30 Sarah Goslee <[hidden email]>

> Hi,
>
> On Fri, Sep 30, 2011 at 5:21 AM, Mathieu Rajerison
> <[hidden email]> wrote:
> > Hello list,
> >
> >
> > I have determined major urban areas.
> >
> > This is just a post to get ideas from R-users on how to qualify urban
> > shapes.
> >
> > The data can either be binary raster (urban/ not urban), either vector.
> >
> > 1) Some urban areas follow linear infrastructures, thus are linear
> > 2) Some other diverge equally from a central heart, and are circular.
> > 3) Some are a mix and are like stars.
>
> That sounds like the kind of task that patch metrics such as perimeter/area
> ratio and fractal dimension were created for. Take a look at the copious
> Fragstats literature. I don't know if any have been implemented in R, but
> wouldn't be surprised.
>
> Sarah
>
> > The idea would be to get an index that give for each area, the
> probability
> > of belonging to each of these three classes. Like a 3-column data frame
> >
> > I wondered if packages already existed, or statistical methods for this
> > purpose. Notably, I think that topographic derivatives derived from
> > smoothed/unsmoothed binary data like aspect, could be used to qualify
> these
> > shapes (?)
> >
> > Thanks for any idea or exchange on the subject!
> >
>
>
> --
> Sarah Goslee
> http://www.functionaldiversity.org
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Sarah Goslee
Mathieu,

On Fri, Sep 30, 2011 at 9:24 AM, Mathieu Rajerison
<[hidden email]> wrote:
> Hi,
>
>
> Maybe you're right: perimeter/ratio could be sufficient...
>
> Fractal dimension and lacunarity are good indicators for urban areas. Do you
> know any R package, tool to quantify these?

I've always just used Fragstats, as I already suggested, but you might look into
SDMtools:
http://cran.r-project.org/web/packages/SDMTools/
http://mo-seph.com/node/302

Sarah

> I have found an ImageJ plugin called Fraclac.
>
> There is another one called r.lacunarity included in spatialtools
> http://www.clusterville.org/spatialtools/index.html
> r.lacunarity is interesting compared to ImageJ::Fraclac because it uses a
> moving window and generates a raster.
>
> I launched a post about lacunarity and fractal dimension on R but didn't
> have any answer.
>
>
> So, if anyone manages to use r.lacunarity or knows other tools than Fraclac,
> I'd be happy!
>
>
> For those interested , here is some literature on fractal dimension applied
> to analysis or aerial images or cities:
> http://www.public.asu.edu/~smyint/publications/CEUS-Lacunarity-Myint-Lam.pdf
> http://cybergeo.revues.org/8902
> http://www.isprs.org/proceedings/XXXVII/congress/3b_pdf/80.pdf
> http://www.fatih.edu.tr/~mcadams/geo352/fractal.pdf
>
>
> 2011/9/30 Sarah Goslee <[hidden email]>
>>
>> Hi,
>>
>> On Fri, Sep 30, 2011 at 5:21 AM, Mathieu Rajerison
>> <[hidden email]> wrote:
>> > Hello list,
>> >
>> >
>> > I have determined major urban areas.
>> >
>> > This is just a post to get ideas from R-users on how to qualify urban
>> > shapes.
>> >
>> > The data can either be binary raster (urban/ not urban), either vector.
>> >
>> > 1) Some urban areas follow linear infrastructures, thus are linear
>> > 2) Some other diverge equally from a central heart, and are circular.
>> > 3) Some are a mix and are like stars.
>>
>> That sounds like the kind of task that patch metrics such as
>> perimeter/area
>> ratio and fractal dimension were created for. Take a look at the copious
>> Fragstats literature. I don't know if any have been implemented in R, but
>> wouldn't be surprised.
>>
>> Sarah
>>
>> > The idea would be to get an index that give for each area, the
>> > probability
>> > of belonging to each of these three classes. Like a 3-column data frame
>> >
>> > I wondered if packages already existed, or statistical methods for this
>> > purpose. Notably, I think that topographic derivatives derived from
>> > smoothed/unsmoothed binary data like aspect, could be used to qualify
>> > these
>> > shapes (?)
>> >
>> > Thanks for any idea or exchange on the subject!
>> >
>>
>>

--
Sarah Goslee
http://www.functionaldiversity.org

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Brian Oney
In reply to this post by Mathieu Rajerison
Hi,
Sarah's idea with FRAGSTATS isn't a bad one. I gather that you would
like to do it in R.
FRAGTSTATS can be used from the command line and anything that can be
accessed with the command line is game for R.
A quick look at the FRAGSTATS manual will tell you that you need to
create  a parameterization file in FRAGSTATS. Then:

system("/Drive:/path_to_Fragstats/Fragstats path_to_file/file.asc
ParameterizationFile.frg///c/")

I hope, I didn't duplicate anything in answering later...

Cheers,
Brian

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Mathieu Rajerison
Hi,

Thanks all for the answers. I'll take a look at FRAGSTATS software. It looks
very interesting.

Mathieu

2011/10/1 Brian Oney <[hidden email]>

> **
> Hi,
> Sarah's idea with FRAGSTATS isn't a bad one. I gather that you would like
> to do it in R.
> FRAGTSTATS can be used from the command line and anything that can be
> accessed with the command line is game for R.
> A quick look at the FRAGSTATS manual will tell you that you need to create
> a parameterization file in FRAGSTATS. Then:
>
> system("*Drive:/path_to_Fragstats/Fragstats path_to_file/file.asc
> ParameterizationFile.frg** /c*")
>
> I hope, I didn't duplicate anything in answering later...
>
> Cheers,
> Brian
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Mathieu Rajerison
For those interested, FRAGSTATS seems to totally answer my issue. Thanks for
making me discover it.

It allows calculation of shape metrics.

http://www.umass.edu/landeco/research/fragstats/documents/Metrics/Shape%20Metrics/SHAPE%20METRICS.htm

I'm not an R developper but I do think it would deserve an R package or at
least an implementation in an existing one :)

2011/10/3 Mathieu Rajerison <[hidden email]>

> Hi,
>
> Thanks all for the answers. I'll take a look at FRAGSTATS software. It
> looks very interesting.
>
> Mathieu
>
>
> 2011/10/1 Brian Oney <[hidden email]>
>
>> **
>> Hi,
>> Sarah's idea with FRAGSTATS isn't a bad one. I gather that you would like
>> to do it in R.
>> FRAGTSTATS can be used from the command line and anything that can be
>> accessed with the command line is game for R.
>> A quick look at the FRAGSTATS manual will tell you that you need to
>> create  a parameterization file in FRAGSTATS. Then:
>>
>> system("*Drive:/path_to_Fragstats/Fragstats path_to_file/file.asc
>> ParameterizationFile.frg** /c*")
>>
>> I hope, I didn't duplicate anything in answering later...
>>
>> Cheers,
>> Brian
>>
>
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Barry Rowlingson
On Mon, Oct 3, 2011 at 10:47 AM, Mathieu Rajerison
<[hidden email]> wrote:

> I'm not an R developper but I do think it would deserve an R package or at
> least an implementation in an existing one :)

 It seems to be only available as a Windows binary download with a
Public Domain license. Step 1 would be asking the authors to publish
their source code under an open source license. Then we can talk about
R packges!

Barry

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Sarah Goslee
Fragstats used to be more freely available; too bad the source version
is no longer available. It hasn't been supported for a very long time,
but used to be downloadable anyway.

Regardless, if I were going to implement that functionality in R, I would
start with the r.le code from GRASS GIS. Using the R-GRASS would
give you almost the same effect, but only if you already use GRASS.

Sarah

On Mon, Oct 3, 2011 at 7:04 AM, Barry Rowlingson
<[hidden email]> wrote:

> On Mon, Oct 3, 2011 at 10:47 AM, Mathieu Rajerison
> <[hidden email]> wrote:
>
>> I'm not an R developper but I do think it would deserve an R package or at
>> least an implementation in an existing one :)
>
>  It seems to be only available as a Windows binary download with a
> Public Domain license. Step 1 would be asking the authors to publish
> their source code under an open source license. Then we can talk about
> R packges!
>
> Barry
>


--
Sarah Goslee
http://www.functionaldiversity.org

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Reply | Threaded
Open this post in threaded view
|

Re: Ideas on qualifying urban shapes: linear / circular / star

Mathieu Rajerison
Otherwise, all the calculations are detailed on this page:
http://www.umass.edu/landeco/research/fragstats/documents/Metrics/Metrics%20TOC.htm

The day I'll go into R developing, it will be my first exercise.


I'm particulary interested in processing a moving window analysis of fractal
dimension, which is a function of perimeter and area. I wondered if it was
possible to calculate the perimeter or a patch of connected cells?

In fact, calculations might be easier to do using vector data instead of
raster and using rgeos library. What do you think?


Best,

Mathieu

2011/10/3 Sarah Goslee <[hidden email]>

> Fragstats used to be more freely available; too bad the source version
> is no longer available. It hasn't been supported for a very long time,
> but used to be downloadable anyway.
>
> Regardless, if I were going to implement that functionality in R, I would
> start with the r.le code from GRASS GIS. Using the R-GRASS would
> give you almost the same effect, but only if you already use GRASS.
>
> Sarah
>
> On Mon, Oct 3, 2011 at 7:04 AM, Barry Rowlingson
> <[hidden email]> wrote:
> > On Mon, Oct 3, 2011 at 10:47 AM, Mathieu Rajerison
> > <[hidden email]> wrote:
> >
> >> I'm not an R developper but I do think it would deserve an R package or
> at
> >> least an implementation in an existing one :)
> >
> >  It seems to be only available as a Windows binary download with a
> > Public Domain license. Step 1 would be asking the authors to publish
> > their source code under an open source license. Then we can talk about
> > R packges!
> >
> > Barry
> >
>
>
> --
> Sarah Goslee
> http://www.functionaldiversity.org
>

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-Geo mailing list
[hidden email]
https://stat.ethz.ch/mailman/listinfo/r-sig-geo