RMS Error

TopGun!
TopGun! Global Mapper User
edited August 2008 in Technical Support
I have two questions:

1.) For georeferencing, I understand the error calculation in GM is the error for each GCP (in Pixels). Does GM have a feature that automatically calculates and displays the total RMS error of all the GCP's?

2.) In general, what GCP error values and total RMS error values are considered acceptable when georeferencing raster maps, or how do I go about finding out what is acceptable? I know this will vary for specific applications, processes, and data.

Thanks,

TopGun!

Comments

  • global_mapper
    global_mapper Administrator
    edited August 2008
    While there is not a way to display the total RMS for all GCPs in Global Mapper, if you export a GCP file from the File menu on the rectification dialog, you could load that into something like Excel and do whatever calculations that you need on the error column, which should be the last one.

    Note that the reported error is in pixels at each control point, so anything under 1 means less than 1 pixel of error which is always acceptable. If you use the default automatic rectification setting the errors for each control point will nearly always be very close to zero.

    Let me know if I can be of further assistance.

    Thanks,

    Mike
    Global Mapper Support
    support@globalmapper.com
  • TopGun!
    TopGun! Global Mapper User
    edited August 2008
    Mike, thanks. Two more Questions:

    Is the rectification process considered acceptable if every GCP's error is less than 1 pixel, or when the calculated RMS error from all the GCP's is less than 1 pixel, or both?

    Could you briefly explain the theory behind where the 1 pixel acceptable/not-acceptable threshold came from?

    Thanks,

    TopGun!
  • global_mapper
    global_mapper Administrator
    edited August 2008
    For precise mapping data you would ideally want the error for all GCPs to be less than 1 pixel. The 1 pixel threshold was posited because the precision of a raster image can never be greater than 1 pixel since that is its smallest individual element, so there is already an inherent error of +/- 1 pixel in a raster image, so a +/- 1 pixel error for a GCP doesn't really change anything.

    Note that so long as you use the Automatic rectification this shouldn't be an issue anyway as the error for each GCP should always be exactly or very close to 0. There was a problem with this for 5 or more GCPs in the latest release (v9.03) that is fixed in the v10 betas and also available as a patched v9 build at http://www.globalmapper.com/global_mapper9.zip.

    Let me know if I can be of further assistance.

    Thanks,

    Mike
    Global Mapper Support
    support@globalmapper.com

Categories

In this Discussion