[Geotiff] Projection CS Encoding

[Pascal Peuch]

I am not sure I totally undestand your question, but 

>>>> It seems as if the tie points always need to be specified in some sort of distance unit, which I don't know and cannot readily compute.  
It is not really a distance unit. It is coordinates in a specific coordinate system. Yours is a projected coordinate system. You have to store TiePoint according to this coordinate system. 
There are no predefined 'units' for values stored in ModelTiePointTag and ModelPixelScaleTag, These values are defined by the coordinate system specified in the GeoKeyDirectoryTag.

>>>>>So, is there any way to encode the information such that the tie points are specified in lat/lon, but still use a Projected CS with the other parameters specified? 
I don't think so. Moreover, I can't figure out why you want to do that?
Should I undestand that you know the tiepoint coordinate only in lat/long and you don't know it in your projected system. If so why just don't you transform it with GEOTRANS utility for example ?


Looks like your image is "rectified". That is : your georeferencing model is the following function. 
X = ax + b
Y = a'x + b'
where (X,Y) are the coordinates of any point in a projection P.
(x,y) are the pixel 'coordinates' of this point
a and a' are derived from pixel size
b and b' are derived from the tiepoint
In your case P is a particular Lambert conformal projection.

The proces is thus :
1/
Look  for your particular lambert projection (depending on standard parallels) in EPSG database.
If it exists : just get the code from EPSG and store it in the adequat geokey
If it doesn't exist, two solutions :
    rectify your image in a widespread projection (at least one existing in EPSG)
    or
    user-defined you projection in geokeys (mind that a lot of readers software are not able to read that type of images)
2/
store in the ModelTiePoint Tag, coordinates (according to your coordinate system) of the upper left pixel (have a look at pixelisArea definition in the specification)
3/
Store the pixel sizes (X and Y spacings in the unit of your coordinate system) in the ModelPixelScale

Hope this will help.

Pascal
  ----- Original Message ----- 
  From: Brent Shaw 
  To: geotiff at remotesensing.org 
  Sent: Thursday, May 12, 2005 10:09 PM
  Subject: [Geotiff] Projection CS Encoding


  Does anyone out there have any experience encoding geotiffs that are referenced with a standard transformation method (e.g., Lambert Conformal, Polar Stereographic, Mercator) on a user-defined projection?  For example, I can create an image where I know:

  1.  Projection is Lambert Conformal 
  2.  I know both standard parallels in degrees
  3.  I know the standard longitude in degrees
  4.  I know the spacing of the pixels along either of the standard latitudes in meters
  5.  I can compute the lat/lon of any raster pixel in the image to use as a tie points.  Since I know items 1-4, I really should only need 1 tie point, but I can compute any of the lat/lons with routines I already have.  

  What I specifically don't understand how to use for this case is the ModelTiePointTag and the PixelScaleTag.  It seems as if the tie points always need to be specified in some sort of distance unit, which I don't know and cannot readily compute.  So, is there anyway to encode the information such that the tie points are specified in lat/lon, but still use a Projected CS with the other parameters specified?  

  Thanks for any help anyone can provide.

  Regards,
  Brent


------------------------------------------------------------------------------


  _______________________________________________
  Geotiff mailing list
  Geotiff at remotesensing.org
  http://xserve.flids.com/mailman/listinfo/geotiff
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/geotiff/attachments/20050514/86d384b3/attachment.html>