3

![old map georeferenced badly][1]

[1]:

I georeferenced this old map using Thin Plate Spline, and Cubic Spline as transformation settings. However, the result was quite bad. Do I have to add points througout the whole rasterfile for this to be even remotely readable, or is the map just too bad to georeference in a good way?

  • 1
    Scanned maps are usually not so much distorted. Second order polynomial or plain affine transformation should give a good result. If not, some ground control points are probably wrongly interpreted. – user30184 Mar 5 '15 at 21:39
  • 1
    How many control points did you use? How spread out were they? You do need some distribution of the points. That said, it does look like a fairly old map which may or may not have its own significant distortions or errors. It also might help in your screenshot of the vector data were on top - as it is, we just see the warped raster with no indication how it's fitting what you're trying to reference to. – Chris W Mar 6 '15 at 1:34
  • I used around 10 points. I guess I need more. Thank you all for your comments! – DOMINUS MIHI ADIUTOR Mar 7 '15 at 15:15
3

For maps that are 100 or more years old, you should not expect too much accuracy. The surveyeors of old times did not have high-precision GPS receivers.

As a consequence, polynomial interpolation is better than thin plate spline, because it averages over all GCPs you enter, while the latter considers the GCPs as exact.

Try to gather as much GCP you can find.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.