Large number of control points

robinrobin Global Mapper UserPosts: 50Trusted User
Imagine you want to simulate NADCON...

Create a grid of points.. Pass it through NADCON.. get output points...

Now, glue them together and create a GCP file.

Then take any dataset, pass it throught NADCON, get new coordinates.

Take same dataset, load into Global mapper, load control points, and select triangulation method.

The transform is not as close as one would like to the NADCON version and I don't think it is an interpolation problem...

Why I ask is that I have come across many situations where affine or even high order polynomials (global mapper only uses second order) are not enough.

In these cases creating coordinate sets and rubersheeting is the only way.

regards,

Robin

Comments

  • global_mapperglobal_mapper Administrator Posts: 17,238
    edited May 2009
    Robin,

    You could do this for some unsupported NADCON or NTv2 data set, but since Global Mapper already uses the NADCON conversions I'm not sure why you would want to do this.

    Thanks,

    Mike
    Global Mapper Support
    support@globalmapper.com
  • robinrobin Global Mapper User Posts: 50Trusted User
    edited May 2009
    We do gogle earth type imagery in areas, where, the existing local datum exceeds the accuracy specification of the project and that data are not integral to the NADCON or NTV2. Thus, those binary grid-shift datasets are insufficient for my needs.

    Of course, if GLOBAL MAPPER could output a NTV2 ot NADCON type file from the control point (GCP) format, then that would be sweeeeeeeeet........

    Then we could create our own little grid file for our clients to use when integtrating high accuracy GPS with local coordinate systems...

    r
  • global_mapperglobal_mapper Administrator Posts: 17,238
    edited May 2009
    There aren't currently any plans to be able to export NTV2 or NADCON files from Global Mapper, but I think there are some 3rd party tools that can do this. I think WinDatum is one I have heard of that can do this. You can get it at http://www2.ogs.trieste.it/Cars/downloads/windatum/InstallWINDATUM.2.3.2.zip.

    Let me know if I can be of further assistance.

    Thanks,

    Mike
    Global Mapper Support
    support@globalmapper.com
  • robinrobin Global Mapper User Posts: 50Trusted User
    edited May 2009
    Hi,

    I figured out what is wrong with the triangulation method... Mike, you clued me in. GM does a peice-wise affine somehow based on the in-out datum points.

    Not sure why you would not have implemented it same way as the elevation tin (internally, you make two tins, one each holds delta-east and delta north respectively) and then interpolate the dE and DN respectivaly, then apply....

    This method, when applied, matches the NADCON and NTV2 methods to a few millimeters.. Of course NADCON used polynomial interploation so perhaps a bit more accurate but millimeters are ok for me...

    I should have clued in when GM asks for 5 points min, instead of 3.... LOL

    Anyway, the other approach to this problem, in fact the preferred one, is to make some additions to the grid shift datum functionality of GM.

    A potential implementation to solve this problem is to put the GCP file in the directory

    C:\Program Files\GlobalMapper10\nadcon...

    but instead of a binary file is is ASCII file which GM already has the function to read! all that remains is: a) to implement the drop down box in the 'add datum' dialog to pick the GCP file in the datums drop down box and b) to interpolate the GCP file a grid with the same internal structure as NTV2 while you reading it... (or perhaps to just store the tin with dE, dN at each vertex and then modify the algo to use plane interpolation)

    I love global mapper. For its simplicity and functionality it is awesome, and I have used (and coded) everything...

    This is one thing that is important because many organizations have there own datum grids, which are not public... not implemented as standard binary grids... and are not part of the GM standard package.... I want to get them to use global mapper because it's simplicity and cost facilitates wide use..

    To make these relatively simple facilities to read in a ASCII grid shift fits very well with the simplicity and scripting philosophies of GM and I hope to see it..

    I will volunteer to code bits of it if necessary! I do c and vb...

    btw, I know the people who wrote the NTV2 standard and implemented the Canadian models..

    Regards,

    Robin
  • robinrobin Global Mapper User Posts: 50Trusted User
    edited May 2009
    ok, if this works and I can create a binary grid shift from a set of randomly distributed ascii points, then can I use the resulting grid shift in GM?
  • global_mapperglobal_mapper Administrator Posts: 17,238
    edited May 2009
    While there is not currently a way to create a custom datum using a grid shift file (either NTv2 or NADCON format or some other ASCII format), this is something that could probably be added pretty easily once you can create the custom grid shift files.

    Regarding the transformation method when using GCP files, the same thing is used during rectification as during elevation grid generation. Basically a TIN is created for the source points to easting and another for source points to northing, then an affine transformation is applied within each triangle of the TIN to sample locations within a triangle. The same thing is done for elevation grids, it's just that a resampling to a regular grid is done for display rather than keeping everything as a TIN and dynamically resampling as is done for rectification.

    If the TIN was resampled to a grid of some spacing and then bilinear interpolation was used on the grid you would basically get the NADCON or NTv2 method. The TIN-based piecewise affine transformation should actually be more accurate since you maintain the full accuracy of the TIN rather than interpolating between some fixed samples, which loses some of the information from the original triangulation. So long as your grid sample spacing is small enough though the difference should be pretty negligible and probably well within the error of your control points anyway.

    Thanks,

    Mike
    Global Mapper Support
    support@globalmapper.com
Sign In or Register to comment.