0

I am georeferencing some old aerial pictures in QGIS. The area encompass mountains, hills and a flat area.

For some pictures the error is quite low and I am satisfied with the overlapping (error is around 2 using the polynomial 2). For others, the error is very high (as much as 100!). What bothers me most, is that I cannot understand the red lines that appear in the screen. When I delate these points with long lines, sometimes the error decreases. Other times the nearby points increase even more the lengths of these lines.

I am using as SR the 3003 and as method for transformation polynomial 1 or 2 or 3 and nearest point. I usually put more than 30 points of control.

  • if you select "Thin Plate Spline" as Transformation Type and "Nearest Neighbour" as Resampling method, your results might improve (visual comparison is need then, though) – sn1ks Aug 7 at 11:48
  • Thank you. Yes, the results have improved. – Geo_it Aug 9 at 15:28
2

These are georeferencing error vectors that visualize the dX and dY values of the georeferencing results. Long lines indicate that there is a large error between where the point is on the ungeoreferenced map and where it lies in reality. This may affect your overall quality of georeferencing. Make sure that the points that you choose cover the whole image and they are placed as evenly as possible.

You can have a look here:

https://docs.qgis.org/2.8/en/docs/training_manual/forestry/map_georeferencing.html

  • Thank you very much for the link. There is something that I can do in order to decrease the error? Usually these errors are appearing in the hilly area. – Geo_it Aug 7 at 8:01
  • Yeah, get newer images, which aren't distorted in hilly areas. Otherwise you can't do much, since this is how the images were made. – Erik Aug 7 at 8:06
  • Unfortunately I have to work with them. Do you think that make a difference using 10 points of control or 30? – Geo_it Aug 7 at 8:11
  • 1
    Do you have a set of predefined ground control points that were measured on the spot? Or can you pick your ground controls points by yourself? If it is the second, then experiment with various ground control points on the hilly areas to achieved a better fit. You can also experiment with the transformation type and re sampling method but please know what you're doing. – vagvaf Aug 7 at 8:12

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.