Georeferening: Cusom datum based on gcp - accuracy at edges

robinrobin Global Mapper UserPosts: 50Trusted User
edited September 2012 in Projection Questions
Hi,

I am using the cusom datum based on gcp list functionality to extend a regional NTV2-type transformation to an exended project area.

I created a list of points and ran it through the nrcan software that reads it. Then, simply make a GCP file out of it.

In one test, I load the GCP file in global mapper and tested it against itself with the function of interest.

By definition, I would think the results shoud be very close or essenitally Identical.

On the points near the very edges, however, I think I may be seeing cases that may indicate significant differences.

Is there information about the handling of the GCP list or the modelling it performs?

Kind Regards,

Robin

P.S. What I have in my GCP file is:

(header line)
myLong, myLat, WGS84Long, WGS84Lat
.
.
.
(about 1000 lines of same)

Comments

  • global_mapperglobal_mapper Administrator Posts: 17,238
    edited September 2012
    Are you using that file to create a new custom datum? If so what happens is that rectification model is created from the input coordinates to the output one. This would use a piecewise-affine (triangulated) rectification for thousands of points like that, which means that the points will be triangulated and a 3-point affine transformation used within each input triangle, so each input GCP should be exactly maintained. However if you convert any points outside of the triangulation you will just get a best-fit polynomial used as the triangulation doesn't really extend outside of the triangulation. Are the points that you are seeing the differences in outside of a triangulation of the points?

    Thanks,

    Mike
    Global Mapper Guru
    gmsupport@bluemarblegeo.com
    http://www.globalmapper.com
  • robinrobin Global Mapper User Posts: 50Trusted User
    edited September 2012
    Hi,

    No, that's the funny thing... the points I am testing are the points in the GCP file! The actual file... not some working version before the GCP was created.... so, we can gurantee that no points are outside the GCP...

    The only thing I can think of is that in the creation of the internal tin, since this set of points in the GCP is a grid, then if there is even 1/10 mm deviation between, say, three points on the edge of the tin, then the possility exists that a very skinny triangle is created...

    Thus, with a near infinitly (actually just very high area-to-perimeter ratio) thin tringle, any difference in the vertex elevatino could exhibit severe slope.

    In the triangulation, or triangulation in general in GM, is it possible to eliminate triangles with only 2 adjacent triangles (external tringle) with a ratio between the perimeter to area ration under a certian threshold to be eliminated? I know it cant do it now but is it impementable?

    Also, what behaviour do you think is good with points outside any tin? Perhaps the best is to not transform them at all and remove from display and exports... if there is a problem, then fix the tin and continue, rather than to export points that are questionable...

    Kind regards,

    Robin
  • global_mapperglobal_mapper Administrator Posts: 17,238
    edited September 2012
    Robin,

    What version of Global Mapper are you using? There did used to be some issues with conversions right at the edge GCPs where round-off would cause the points to be just very slightly outside of the triangle. I think in the latest versions though there is a bit of fudge to account for that case though.

    Thanks,

    Mike
    Global Mapper Guru
    gmsupport@bluemarblegeo.com
    http://www.globalmapper.com
Sign In or Register to comment.