1

I have several maps to digitize. The base maps were Canadian NTS map sheets, all for more or less the same region. The useful information has been traced on top with handwritten markings. The maps were large so they had to be photographed rather than scanned.

It seems to me that I should be able to re-use ground control points to some extent after georeferencing the first map. Certain internal features on each map should be the same relative distance from each other.

What I want to do is manually georeference map A, and then warp map A's ground points onto map B and correct their positions. It seems to me that this would save time compared to manually entering all new control points - clearly there will be some error in the transform, but if control points are near enough to their true positions I can just shift them into place instead re-entering them.

This should because the two maps are in the same projection. The images of the base maps are same subject to translation, scaling, rotation, and some unavoidable error.

So this is a lot like re-projection, I just want to reproject a set of control points instead of using control points to re-project a raster image.

I'm using QGIS. Is there a way to accomplish this in QGIS?

6
  • qgistutorials.com/en/docs/georeferencing_basics.html
    – Mapperz
    Commented Mar 8, 2015 at 21:25
  • @Mapperz: What I'm trying to do is not "basic", but thanks.
    – jbarlow
    Commented Mar 8, 2015 at 21:29
  • These maps are available via wms gis.stackexchange.com/questions/4298/…
    – Mapperz
    Commented Mar 8, 2015 at 21:33
  • @Mapperz: The maps have handwritten marks which are the data I need from them. I already have those base maps. After I set up GCPs on the images I would reproject them using those vector maps as a guide.
    – jbarlow
    Commented Mar 8, 2015 at 21:35
  • So basically you have the same control points you're referencing to and in all of the images, but their coordinates in each image varies? In theory you could do what you say - reference A, save out the link table, load it in to B and then move/adjust the image coordinate of the link pair to the correct spot. I'm not sure that would be any faster than just having a set of points you could snap to and creating all new links though.
    – Chris W
    Commented Mar 8, 2015 at 21:42

1 Answer 1

2

I learned that ArcGIS has a georeferencing feature called autoregistration that performs this task with some caveats. They say it does not work well in my use case of old scanned maps.

QGIS has no similar feature, so I wrote a script in Python using the OpenCV library that looks for general image similarity, maps the correspondence, and then transfers points from a georeferenced point list to equivalent positions on the other image. Results were very good - it saved me weeks of work. (Unfortunately I can't open source it since it was developed on company time, but you can PM if interested.)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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