i.rectify2

Brian Buckley buckley at ig0.rs.msu.edu
Fri Jan 14 15:20:49 EST 1994


> Hello  all,
> 
> We have rasterdata in lat-lon which cover a large part of Europe (From Scandinavia to South-Europe; from the UK to east Russia). We need these data in a mapset with lcc coordinates.
> I rectified the maps with i.rectify2 (3rd order). I edited the POINTS-file for which I calculated the lcc coordinates for 15 points,distributed evenly over the area, exactly with m.proj.
> The results are, however too inaccurate (Especially in the corners: the British isles, the Greek islands and southern Italy). Now my question:
> Has anyone got a solution to increase the accuracy?
> It is a pitty that r.in.ll is not made for the other projections, but only for UTM.
> 
> --------------------------------------------
> ps. proj.info:
> 
> 
> name: Lambert Conformal Conic
> proj: lcc
> ellps: international
> a: 6378388.0000000000
> es: 0.0067226700
> lat_0: 52.0000000000
> lon_0: 6.0000000000
> lat_1: 33.0000000000
> lat_2: 70.0000000000
> ------------------------------------------------
> 
> Thanks, Harm olthof
> 

Harm,
I wrote the i.rectify2 program (by modifying the i.rectify program).  i.rectify
can only perform a first order polynomial transformation, while i.rectify2 can
do a first, second or third order polynomial transformation.  i.rectify2,
which is a general transformation program to be used when there is no
mathematical formula available to use, should not be confused with the lat/lon
to Lambert conformal conic projection transformation which is a precise, and
complex, mathematical formula.  To try to emulate the ll -> lcc transformation
with even a third order polynomial transformation is imprecise at best and
will become quite poor and pretty much unusable as the geographic area becomes
large.  i.rectify2 was written to help rectify ariel photography where there
is no other way to rectify the photo.
Someone mention an r.proj program which sounds like that will be your best
bet, but if you do need to use i.rectify2, keep your regions as small as
possible to minimize the error to something that you can live with.  If your
areas are so small that you have trouble getting the 10 control points needed
for 3rd order transformations, we have had relatively good sucess rectifying
ariel photos with 2nd order transformations which only require 6 control
points.  We have also noticed substantial "side effects" with 3rd order
transformations with close to the minimum number of control points (10,11 or
12) or poorly located control points, so in general, don't use a higher order
than you really need to suit your purpose and give you the accuracy that you
need.
Hope this helped,
:-) Brian
+-------------------------------------------------------------+
| Brian J. Buckley                Michigan State University   |
| Systems Analyst                 Center for Remote Sensing   |
| E-Mail: buckley at ig0.rs.msu.edu  115 Manly Miles Building    |
| Phone: (517)355-4519            1405 South Harrison Road    |
| Fax: (517)353-1821              East Lansing, MI 48824  USA |
+-------------------------------------------------------------+





More information about the grass-user mailing list