On Sat, 16 Jun 2007, David Hugh-Jones wrote:
> Thank you for this help. I realised after writing this that indeed,
> spatstat won't work because I have world-scale data in unprojected
> latitude/longitude form... so my distance measure is no use. This
> leads to more questions
>
> - is there a canonical way to convert projections from within "sp"? My
> current thought is to get the raw data out and use "mapproj".
The spTransform() methods in rgdal.
> - when that's done, is there a function to measure great circle distance?
Well, if your data are world data, you probably want to work in great
circle distances directly, which you can get from spDistsN1 from a matrix
of points to a single point.
> - would the algorithm "measure distance between all points of A and
> all lines of B, then the same with A and B reversed, and take the
> minimum", still work? I can't see why not but my intuition is not
> clear.
With GC distances, point to line is much harder. You may be obliged to use
point to point, which will be symmetric, but is an approximation of line
to line. If the differences may matter badly, you'll need to interpolate
points and add them to long straight segments.
It may also be worth asking on the PostGIS list whether you could upload
your geometries to PostGIS and have GEOS do the measurement - if you have
access to the software and experience with it. It may be that it can
handle geographical coordinates.
Best wishes,
Roger
>
> Cheers
> David
>
>
> On 16/06/07, Roger Bivand <[EMAIL PROTECTED]> wrote:
>> On Fri, 15 Jun 2007, David Hugh-Jones wrote:
>>
>>> Hi R-spatialists,
>>>
>>> I'm working with a big list of polygons (imported into R with package
>>> sp as a SpatialPolygonDataFrame, and I am trying to calculate minimum
>>> distances between each pair of polygons. For this, I need to get the
>>> lines for each polygons. At the moment I can't seem to find a way to
>>> do this except to delve right in with
>>>
>>> for (p in [EMAIL PROTECTED]) {
>>> for (q in [EMAIL PROTECTED])
>>> linesc <- rbind(linesc, coordinates(q))
>>> }
>>>
>>> and then to add the endpoints of linesc as columns 3 and 4 of the
>>> matrix. Is there a simpler way to extract lines from polygons?
>>
>> No, this looks OK, though I would have said slot(spdfc, "polygons"), but
>> there is no difference in practice. I would have kept the coordinates in a
>> list of matrices rather than one matrix, in the first loop, because here
>> you lose the polygon IDs.
>>
>> You could go the way you are going now, then use the psp() and crossdist()
>> functions in the spatstat package - they take single line segment
>> patterns, so you'd need to break out the coordinates further to make psp
>> objects, probably all in one shot.
>>
>> I'd get the full matrix of distances, and retrieve the proper triangle by
>> comparing pairs, which are not guaranteed to be equal - see the crossdist
>> help page.
>>
>> If you have geographical coordinates and world scale data, meaning you
>> have to use Great Circle distances, the route via spapstat probably isn't
>> viable. If you can project, you'd do that first, of course.
>>
>> This is an interesting question, and I'm sure others would value hearing
>> about your solution.
>>
>> Best wishes,
>>
>> Roger
>>
>>>
>>> Cheers
>>> David
>>>
>>> _______________________________________________
>>> R-sig-Geo mailing list
>>> R-sig-Geo@stat.math.ethz.ch
>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>
>>
>> --
>> Roger Bivand
>> Economic Geography Section, Department of Economics, Norwegian School of
>> Economics and Business Administration, Helleveien 30, N-5045 Bergen,
>> Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
>> e-mail: [EMAIL PROTECTED]
>>
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo@stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>
--
Roger Bivand
Economic Geography Section, Department of Economics, Norwegian School of
Economics and Business Administration, Helleveien 30, N-5045 Bergen,
Norway. voice: +47 55 95 93 55; fax +47 55 95 95 43
e-mail: [EMAIL PROTECTED]
_______________________________________________
R-sig-Geo mailing list
R-sig-Geo@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
